Composition rankings

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Despoiler: damage resolution is fully connected in combat. The older one-sided base-DPS preview lacks an enemy Health pool and can apply only the base amplification.

Akali: target-selection details when the AP form recasts

Varus: Official ability variables are unavailable. The split of 385/580/925% AD plus 30/45/70% AP is consistent across two sources but remains provisional.; The projectile uses an inferred half-width of 0.65 hexes and a range of 10 hexes so the arrow can leave the board. Official text describes a straight arrow without specifying range or exact geometry.; Edited video cannot establish exact charge timing. Cast windup and recovery still use the engine's provisional defaults.

Shen: Shen's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Nearby has no official world-space radius. The model explicitly assumes 3 hexes pending official data or repeatable board tests.; Official tie-breaking among equally close injured allies and eligibility of minions or summons remain unverified. The engine currently excludes dummies and relics while allowing real champions.; The video verifies two shields and a subsequent empowered attack only. Current text says bonus magic damage, so the model adds it to the base attack rather than replacing that attack.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Diana: Diana's official spell data is unavailable. Six orbs, 2 hexes, damage per orb [65,100,155], shields [150,225,300] and 2 seconds agree across two sources but remain provisional.; Sources do not specify whether six orbs are distributed evenly, randomly or by other weights among valid enemies. The model assigns each orb independently and records the decisions; this is not an official selection rule.; A clear current-version video, precise projectile travel, behavior after death before impact and independent expiry of multiple shields are unverified. The model uses a default 0.3-second impact delay and shared shield expiry.

Rakan: Rakan's official spell values are unavailable. The 18.2 shields and Attack Speed values by star level come from the same-day tft.tools snapshot and remain provisional.; The local Lolchess-derived snapshot and current tft.tools agree on 1- to 3-star shields and 35/105 Mana. The 4-star shield of 490 appears only on tft.tools. Provisional spell values remain in the specialization layer and do not replace unverified official base stats.; The movement shown at 41 seconds in the earlier preview video conflicts with the current 18.2 text and may precede a rework or be mislabeled. It is not treated as evidence of the current ability.; If the ally with the highest damage dies before impact, the model compares surviving allies again. Whether Rakan can target himself, official tie-breaking, and whether the shield and Attack Speed apply on the same server tick remain unverified.

9 units · 1 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Despoiler: damage resolution is fully connected in combat. The older one-sided base-DPS preview lacks an enemy Health pool and can apply only the base amplification.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Shen: Shen's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Nearby has no official world-space radius. The model explicitly assumes 3 hexes pending official data or repeatable board tests.; Official tie-breaking among equally close injured allies and eligibility of minions or summons remain unverified. The engine currently excludes dummies and relics while allowing real champions.; The video verifies two shields and a subsequent empowered attack only. Current text says bonus magic damage, so the model adds it to the base attack rather than replacing that attack.

Akali: target-selection details when the AP form recasts

Varus: Official ability variables are unavailable. The split of 385/580/925% AD plus 30/45/70% AP is consistent across two sources but remains provisional.; The projectile uses an inferred half-width of 0.65 hexes and a range of 10 hexes so the arrow can leave the board. Official text describes a straight arrow without specifying range or exact geometry.; Edited video cannot establish exact charge timing. Cast windup and recovery still use the engine's provisional defaults.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Diana: Diana's official spell data is unavailable. Six orbs, 2 hexes, damage per orb [65,100,155], shields [150,225,300] and 2 seconds agree across two sources but remain provisional.; Sources do not specify whether six orbs are distributed evenly, randomly or by other weights among valid enemies. The model assigns each orb independently and records the decisions; this is not an official selection rule.; A clear current-version video, precise projectile travel, behavior after death before impact and independent expiry of multiple shields are unverified. The model uses a default 0.3-second impact delay and shared shield expiry.

Rakan: Rakan's official spell values are unavailable. The 18.2 shields and Attack Speed values by star level come from the same-day tft.tools snapshot and remain provisional.; The local Lolchess-derived snapshot and current tft.tools agree on 1- to 3-star shields and 35/105 Mana. The 4-star shield of 490 appears only on tft.tools. Provisional spell values remain in the specialization layer and do not replace unverified official base stats.; The movement shown at 41 seconds in the earlier preview video conflicts with the current 18.2 text and may precede a rework or be mislabeled. It is not treated as evidence of the current ability.; If the ally with the highest damage dies before impact, the model compares surviving allies again. Whether Rakan can target himself, official tie-breaking, and whether the shield and Attack Speed apply on the same server tick remain unverified.

8 units · 2 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Despoiler: damage resolution is fully connected in combat. The older one-sided base-DPS preview lacks an enemy Health pool and can apply only the base amplification.

Akali: target-selection details when the AP form recasts

Varus: Official ability variables are unavailable. The split of 385/580/925% AD plus 30/45/70% AP is consistent across two sources but remains provisional.; The projectile uses an inferred half-width of 0.65 hexes and a range of 10 hexes so the arrow can leave the board. Official text describes a straight arrow without specifying range or exact geometry.; Edited video cannot establish exact charge timing. Cast windup and recovery still use the engine's provisional defaults.

Shen: Shen's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Nearby has no official world-space radius. The model explicitly assumes 3 hexes pending official data or repeatable board tests.; Official tie-breaking among equally close injured allies and eligibility of minions or summons remain unverified. The engine currently excludes dummies and relics while allowing real champions.; The video verifies two shields and a subsequent empowered attack only. Current text says bonus magic damage, so the model adds it to the base attack rather than replacing that attack.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Diana: Diana's official spell data is unavailable. Six orbs, 2 hexes, damage per orb [65,100,155], shields [150,225,300] and 2 seconds agree across two sources but remain provisional.; Sources do not specify whether six orbs are distributed evenly, randomly or by other weights among valid enemies. The model assigns each orb independently and records the decisions; this is not an official selection rule.; A clear current-version video, precise projectile travel, behavior after death before impact and independent expiry of multiple shields are unverified. The model uses a default 0.3-second impact delay and shared shield expiry.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Solar: opening shields, shield and bonus magic-damage growth per unique 3-star ally, team Attack Speed/resistances at three 3-stars and conversion of 40% of bonus magic damage to true damage at five are implemented. Randomly upgrading a 3-star to 4-star every 4 seconds after reaching eight 3-stars is missing because 4-star ability data remains incomplete.

Leona: Leona's official Set 18 spell data is unavailable. Starting resistances, Armor scaling and stun values remain provisional, cross-checked between Lolchess and TFTips.; The video confirms the visible decay, target and stun but cannot establish the precise AP snapshot time. The model uses the usual combat-start snapshot.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Kayle: Direction, area and maximum target count of the 3-star attack wave; Official ordering of Magic Resist reduction and bonus magic damage within one attack

Sejuani: Sejuani's official spell data is unavailable. Shields, the 10% maximum Health component, cone damage and line damage by star level use provisional values cross-checked between Lolchess and TFTips.; Video and text confirm a cone and a line but not precisely a 90-degree angle, 3-hex range, 0.5-hex half-width or 5-target cap. These remain provisional geometry pending official verification.; The delay between stages, reorientation after the target dies, shield-refresh merging and whether the line starts at the cone's end are unknown. The model uses 0.1 seconds, locks direction on cast, adds shields normally and uses the caster as the origin for both stages.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Alune: Official data has not populated NumMoonshards, NumEnemies or MagicDamageCalc1/2.; Normal and full-moon coefficients by star level remain provisional.; The available direct binary payload conflicts with the current-set ability and cannot be promoted to verified data.

Fiddlesticks: The available binary data confirms only a 0.25-second cast and base champion data. The nearest three targets, 10 flat Magic Resist, 2 seconds, healing [410,485,850] and magic damage [70,105,170] remain provisional.; The video confirms three simultaneous visible tethers, periodic damage to multiple targets and periodic self-healing. It does not verify 10 Magic Resist, 2 seconds, star-scaled coefficients, server tick frequency or whether total healing changes with surviving targets.; The 0.5-second interval is a damage-conserving calculation convention, not a verified server frequency. One total self-healing pool per cast follows the third-party singular healing wording and still needs official verification.; Sources do not specify whether the drain locks normal attacks, can be interrupted by control, or ends individual tethers when targets leave range or die. The model schedules persistent events on locked targets without inventing extra channel-interruption rules.

Ahri: Official ability data is unavailable. Lolchess and TFTips agree on 485/735/3500; the full set of 1.75 seconds, 3 hexes, 20% and the 20% minimum is mainly from TFTips and remains provisional.; Third-party text does not reveal tie-breaking between clusters with equal hit counts. The model chooses the closer cluster, then uses stable row/column ordering pending official verification.; The video does not establish whether the blast center is selected at cast start or impact. The model locks it on cast start rather than automatically tracking a new cluster at impact.

Yorick: CommunityDragon has character data for Yorick and his Spirit Walker, but the aggregate Set 18 snapshot contained only two shop champions. Active damage, healing and summon Health formulas still use provisional Lolchess values pending verification.; Spirit Walker data does not provide usable base AD, Armor or role classification. Unknown AD and Armor remain null in the source and use a neutral zero in combat; they are not inferred from the video's 27 damage.; The video confirms the order of the single-target active and the summon on death. Full summon attack rules, lifetime, inherited effects and retargeting when the target dies before impact remain unverified.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Solar: opening shields, shield and bonus magic-damage growth per unique 3-star ally, team Attack Speed/resistances at three 3-stars and conversion of 40% of bonus magic damage to true damage at five are implemented. Randomly upgrading a 3-star to 4-star every 4 seconds after reaching eight 3-stars is missing because 4-star ability data remains incomplete.

Kayle: Direction, area and maximum target count of the 3-star attack wave; Official ordering of Magic Resist reduction and bonus magic damage within one attack

Sejuani: Sejuani's official spell data is unavailable. Shields, the 10% maximum Health component, cone damage and line damage by star level use provisional values cross-checked between Lolchess and TFTips.; Video and text confirm a cone and a line but not precisely a 90-degree angle, 3-hex range, 0.5-hex half-width or 5-target cap. These remain provisional geometry pending official verification.; The delay between stages, reorientation after the target dies, shield-refresh merging and whether the line starts at the cone's end are unknown. The model uses 0.1 seconds, locks direction on cast, adds shields normally and uses the caster as the origin for both stages.

Leona: Leona's official Set 18 spell data is unavailable. Starting resistances, Armor scaling and stun values remain provisional, cross-checked between Lolchess and TFTips.; The video confirms the visible decay, target and stun but cannot establish the precise AP snapshot time. The model uses the usual combat-start snapshot.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Alune: Official data has not populated NumMoonshards, NumEnemies or MagicDamageCalc1/2.; Normal and full-moon coefficients by star level remain provisional.; The available direct binary payload conflicts with the current-set ability and cannot be promoted to verified data.

Fiddlesticks: The available binary data confirms only a 0.25-second cast and base champion data. The nearest three targets, 10 flat Magic Resist, 2 seconds, healing [410,485,850] and magic damage [70,105,170] remain provisional.; The video confirms three simultaneous visible tethers, periodic damage to multiple targets and periodic self-healing. It does not verify 10 Magic Resist, 2 seconds, star-scaled coefficients, server tick frequency or whether total healing changes with surviving targets.; The 0.5-second interval is a damage-conserving calculation convention, not a verified server frequency. One total self-healing pool per cast follows the third-party singular healing wording and still needs official verification.; Sources do not specify whether the drain locks normal attacks, can be interrupted by control, or ends individual tethers when targets leave range or die. The model schedules persistent events on locked targets without inventing extra channel-interruption rules.

Ahri: Official ability data is unavailable. Lolchess and TFTips agree on 485/735/3500; the full set of 1.75 seconds, 3 hexes, 20% and the 20% minimum is mainly from TFTips and remains provisional.; Third-party text does not reveal tie-breaking between clusters with equal hit counts. The model chooses the closer cluster, then uses stable row/column ordering pending official verification.; The video does not establish whether the blast center is selected at cast start or impact. The model locks it on cast start rather than automatically tracking a new cluster at impact.

Yorick: CommunityDragon has character data for Yorick and his Spirit Walker, but the aggregate Set 18 snapshot contained only two shop champions. Active damage, healing and summon Health formulas still use provisional Lolchess values pending verification.; Spirit Walker data does not provide usable base AD, Armor or role classification. Unknown AD and Armor remain null in the source and use a neutral zero in combat; they are not inferred from the video's 27 damage.; The video confirms the order of the single-target active and the summon on death. Full summon attack rules, lifetime, inherited effects and retargeting when the target dies before impact remain unverified.

7 units · 2 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Akali: target-selection details when the AP form recasts

Varus: Official ability variables are unavailable. The split of 385/580/925% AD plus 30/45/70% AP is consistent across two sources but remains provisional.; The projectile uses an inferred half-width of 0.65 hexes and a range of 10 hexes so the arrow can leave the board. Official text describes a straight arrow without specifying range or exact geometry.; Edited video cannot establish exact charge timing. Cast windup and recovery still use the engine's provisional defaults.

Shen: Shen's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Nearby has no official world-space radius. The model explicitly assumes 3 hexes pending official data or repeatable board tests.; Official tie-breaking among equally close injured allies and eligibility of minions or summons remain unverified. The engine currently excludes dummies and relics while allowing real champions.; The video verifies two shields and a subsequent empowered attack only. Current text says bonus magic damage, so the model adds it to the base attack rather than replacing that attack.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

8 units · 1 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Despoiler: damage resolution is fully connected in combat. The older one-sided base-DPS preview lacks an enemy Health pool and can apply only the base amplification.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Shen: Shen's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Nearby has no official world-space radius. The model explicitly assumes 3 hexes pending official data or repeatable board tests.; Official tie-breaking among equally close injured allies and eligibility of minions or summons remain unverified. The engine currently excludes dummies and relics while allowing real champions.; The video verifies two shields and a subsequent empowered attack only. Current text says bonus magic damage, so the model adds it to the base attack rather than replacing that attack.

Akali: target-selection details when the AP form recasts

Varus: Official ability variables are unavailable. The split of 385/580/925% AD plus 30/45/70% AP is consistent across two sources but remains provisional.; The projectile uses an inferred half-width of 0.65 hexes and a range of 10 hexes so the arrow can leave the board. Official text describes a straight arrow without specifying range or exact geometry.; Edited video cannot establish exact charge timing. Cast windup and recovery still use the engine's provisional defaults.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Diana: Diana's official spell data is unavailable. Six orbs, 2 hexes, damage per orb [65,100,155], shields [150,225,300] and 2 seconds agree across two sources but remain provisional.; Sources do not specify whether six orbs are distributed evenly, randomly or by other weights among valid enemies. The model assigns each orb independently and records the decisions; this is not an official selection rule.; A clear current-version video, precise projectile travel, behavior after death before impact and independent expiry of multiple shields are unverified. The model uses a default 0.3-second impact delay and shared shield expiry.

8 units · 2 stars · 0 augments

Primal: — · Pixies 0

Solar: opening shields, shield and bonus magic-damage growth per unique 3-star ally, team Attack Speed/resistances at three 3-stars and conversion of 40% of bonus magic damage to true damage at five are implemented. Randomly upgrading a 3-star to 4-star every 4 seconds after reaching eight 3-stars is missing because 4-star ability data remains incomplete.

Kayle: Direction, area and maximum target count of the 3-star attack wave; Official ordering of Magic Resist reduction and bonus magic damage within one attack

Sejuani: Sejuani's official spell data is unavailable. Shields, the 10% maximum Health component, cone damage and line damage by star level use provisional values cross-checked between Lolchess and TFTips.; Video and text confirm a cone and a line but not precisely a 90-degree angle, 3-hex range, 0.5-hex half-width or 5-target cap. These remain provisional geometry pending official verification.; The delay between stages, reorientation after the target dies, shield-refresh merging and whether the line starts at the cone's end are unknown. The model uses 0.1 seconds, locks direction on cast, adds shields normally and uses the caster as the origin for both stages.

Leona: Leona's official Set 18 spell data is unavailable. Starting resistances, Armor scaling and stun values remain provisional, cross-checked between Lolchess and TFTips.; The video confirms the visible decay, target and stun but cannot establish the precise AP snapshot time. The model uses the usual combat-start snapshot.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Alune: Official data has not populated NumMoonshards, NumEnemies or MagicDamageCalc1/2.; Normal and full-moon coefficients by star level remain provisional.; The available direct binary payload conflicts with the current-set ability and cannot be promoted to verified data.

Fiddlesticks: The available binary data confirms only a 0.25-second cast and base champion data. The nearest three targets, 10 flat Magic Resist, 2 seconds, healing [410,485,850] and magic damage [70,105,170] remain provisional.; The video confirms three simultaneous visible tethers, periodic damage to multiple targets and periodic self-healing. It does not verify 10 Magic Resist, 2 seconds, star-scaled coefficients, server tick frequency or whether total healing changes with surviving targets.; The 0.5-second interval is a damage-conserving calculation convention, not a verified server frequency. One total self-healing pool per cast follows the third-party singular healing wording and still needs official verification.; Sources do not specify whether the drain locks normal attacks, can be interrupted by control, or ends individual tethers when targets leave range or die. The model schedules persistent events on locked targets without inventing extra channel-interruption rules.

Ahri: Official ability data is unavailable. Lolchess and TFTips agree on 485/735/3500; the full set of 1.75 seconds, 3 hexes, 20% and the 20% minimum is mainly from TFTips and remains provisional.; Third-party text does not reveal tie-breaking between clusters with equal hit counts. The model chooses the closer cluster, then uses stable row/column ordering pending official verification.; The video does not establish whether the blast center is selected at cast start or impact. The model locks it on cast start rather than automatically tracking a new cluster at impact.

8 units · 2 stars · 0 augments

Primal: — · Pixies 0

Solar: opening shields, shield and bonus magic-damage growth per unique 3-star ally, team Attack Speed/resistances at three 3-stars and conversion of 40% of bonus magic damage to true damage at five are implemented. Randomly upgrading a 3-star to 4-star every 4 seconds after reaching eight 3-stars is missing because 4-star ability data remains incomplete.

Leona: Leona's official Set 18 spell data is unavailable. Starting resistances, Armor scaling and stun values remain provisional, cross-checked between Lolchess and TFTips.; The video confirms the visible decay, target and stun but cannot establish the precise AP snapshot time. The model uses the usual combat-start snapshot.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Kayle: Direction, area and maximum target count of the 3-star attack wave; Official ordering of Magic Resist reduction and bonus magic damage within one attack

Sejuani: Sejuani's official spell data is unavailable. Shields, the 10% maximum Health component, cone damage and line damage by star level use provisional values cross-checked between Lolchess and TFTips.; Video and text confirm a cone and a line but not precisely a 90-degree angle, 3-hex range, 0.5-hex half-width or 5-target cap. These remain provisional geometry pending official verification.; The delay between stages, reorientation after the target dies, shield-refresh merging and whether the line starts at the cone's end are unknown. The model uses 0.1 seconds, locks direction on cast, adds shields normally and uses the caster as the origin for both stages.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Alune: Official data has not populated NumMoonshards, NumEnemies or MagicDamageCalc1/2.; Normal and full-moon coefficients by star level remain provisional.; The available direct binary payload conflicts with the current-set ability and cannot be promoted to verified data.

Fiddlesticks: The available binary data confirms only a 0.25-second cast and base champion data. The nearest three targets, 10 flat Magic Resist, 2 seconds, healing [410,485,850] and magic damage [70,105,170] remain provisional.; The video confirms three simultaneous visible tethers, periodic damage to multiple targets and periodic self-healing. It does not verify 10 Magic Resist, 2 seconds, star-scaled coefficients, server tick frequency or whether total healing changes with surviving targets.; The 0.5-second interval is a damage-conserving calculation convention, not a verified server frequency. One total self-healing pool per cast follows the third-party singular healing wording and still needs official verification.; Sources do not specify whether the drain locks normal attacks, can be interrupted by control, or ends individual tethers when targets leave range or die. The model schedules persistent events on locked targets without inventing extra channel-interruption rules.

Ahri: Official ability data is unavailable. Lolchess and TFTips agree on 485/735/3500; the full set of 1.75 seconds, 3 hexes, 20% and the 20% minimum is mainly from TFTips and remains provisional.; Third-party text does not reveal tie-breaking between clusters with equal hit counts. The model chooses the closer cluster, then uses stable row/column ordering pending official verification.; The video does not establish whether the blast center is selected at cast start or impact. The model locks it on cast start rather than automatically tracking a new cluster at impact.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Rivals: With Ahead of the Curve active, the latest data confirms the link from Kha'Zix Mana restoration to Rengar at 50%, and from Rengar's actual self-healing to Kha'Zix at 100%. Persistent takedown counts for a lone Rival and the additional 3 progress for killing the other Rival are not connected.

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Elderwood: three separate combat entities, the total-star panel, death stun, periodic AD/AP, Deepwood damage and healing/Attack Speed on ally death are connected. Values use the September 5, 2026 TFTips snapshot. Additional rules at 11 units, secondary-hit geometry and healing scaling need official calibration.

Kha'Zix: Official ability data is unavailable. The two damage sets, 285/400/580 and 310/445/660, and their conditions agree across two sources but remain provisional.; The video cannot establish landing-cell selection or tie-breaking. The model selects the nearest empty adjacent cell to the caster, breaks ties consistently by row/column and uses a default 0.3-second cast pending official verification.; Official cancellation, tracking or in-place resolution is unknown when the target dies, becomes untargetable or has no valid empty adjacent cell before landing. The model cancels without redirecting to another target.

Rengar: Official spell data is unavailable. Damage and minimum/maximum healing use same-day provisional third-party values, with differences between older TFTips/tft.tools snapshots and newer sources.; The description says healing increases with the target's missing Health up to a maximum but provides no curve. The model linearly interpolates from minimum at 0% missing Health to maximum at 100%; this is not an official curve.; The video cannot establish the jump path, exact landing cell or windup/recovery. The model uses the nearest available cell and a 0.2-second hop.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Ezreal: Ezreal's official spell values are unavailable. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Blink away has no exact published distance. The model allows up to 2 hexes and selects an available cell that maximizes distance from the nearest enemy, pending official or measurable video evidence.; The current English text does not say instead. The fourth cast therefore keeps its primary single-target damage and adds the wave. If official scripts show replacement behavior, the normal hit must be skipped.; The 8-hex range, 0.65-hex half-width and preference for the current target when hit counts tie are provisional geometry and tie-breaking, not official specifications.

Gnar: Official leap selection, arc, collision half-width and exact landing timing lack binary data. The model uses legal landing cells within 3 hexes and approximates line collision from the thrown unit's origin to the farthest enemy.; Star-scaled coefficients use cross-checked provisional values and must not be labeled official exact values.

Alistar: Alistar's official spell data is unavailable. The three star-scaled value sets and 8% maximum Health component use provisional values cross-checked between Lolchess and TFTips.; The video caption says knockup, while current Lolchess text says a 1.5-second stun. The model follows the current text and does not label it a knockup; official verification is still required.; The third-party phrase all impairing effects has no official enumeration. The model clears external crowd control only, without also clearing Burn, Wound, Attack Speed reduction or Stasis. Retargeting if the current target dies before impact remains unverified.

Kobuko: Official data does not specify the server interval for continuous healing. The model's four ticks at 0.5-second intervals are a damage-conserving calculation convention.; The binary data provides castFrame=22 without a directly usable cast time. The model retains a provisional 0.3-second impact time.; Official text and binary data do not specify crit eligibility for the replacement magic attack, whether recasts refresh or stack charges, or when maximum Health is sampled. The model allows normal attack crits, refreshes to one charge and samples maximum Health when the attack launches.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Rivals: With Ahead of the Curve active, the latest data confirms the link from Kha'Zix Mana restoration to Rengar at 50%, and from Rengar's actual self-healing to Kha'Zix at 100%. Persistent takedown counts for a lone Rival and the additional 3 progress for killing the other Rival are not connected.

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Elderwood: three separate combat entities, the total-star panel, death stun, periodic AD/AP, Deepwood damage and healing/Attack Speed on ally death are connected. Values use the September 5, 2026 TFTips snapshot. Additional rules at 11 units, secondary-hit geometry and healing scaling need official calibration.

Kha'Zix: Official ability data is unavailable. The two damage sets, 285/400/580 and 310/445/660, and their conditions agree across two sources but remain provisional.; The video cannot establish landing-cell selection or tie-breaking. The model selects the nearest empty adjacent cell to the caster, breaks ties consistently by row/column and uses a default 0.3-second cast pending official verification.; Official cancellation, tracking or in-place resolution is unknown when the target dies, becomes untargetable or has no valid empty adjacent cell before landing. The model cancels without redirecting to another target.

Rengar: Official spell data is unavailable. Damage and minimum/maximum healing use same-day provisional third-party values, with differences between older TFTips/tft.tools snapshots and newer sources.; The description says healing increases with the target's missing Health up to a maximum but provides no curve. The model linearly interpolates from minimum at 0% missing Health to maximum at 100%; this is not an official curve.; The video cannot establish the jump path, exact landing cell or windup/recovery. The model uses the nearest available cell and a 0.2-second hop.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Ezreal: Ezreal's official spell values are unavailable. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Blink away has no exact published distance. The model allows up to 2 hexes and selects an available cell that maximizes distance from the nearest enemy, pending official or measurable video evidence.; The current English text does not say instead. The fourth cast therefore keeps its primary single-target damage and adds the wave. If official scripts show replacement behavior, the normal hit must be skipped.; The 8-hex range, 0.65-hex half-width and preference for the current target when hit counts tie are provisional geometry and tie-breaking, not official specifications.

Gnar: Official leap selection, arc, collision half-width and exact landing timing lack binary data. The model uses legal landing cells within 3 hexes and approximates line collision from the thrown unit's origin to the farthest enemy.; Star-scaled coefficients use cross-checked provisional values and must not be labeled official exact values.

Alistar: Alistar's official spell data is unavailable. The three star-scaled value sets and 8% maximum Health component use provisional values cross-checked between Lolchess and TFTips.; The video caption says knockup, while current Lolchess text says a 1.5-second stun. The model follows the current text and does not label it a knockup; official verification is still required.; The third-party phrase all impairing effects has no official enumeration. The model clears external crowd control only, without also clearing Burn, Wound, Attack Speed reduction or Stasis. Retargeting if the current target dies before impact remains unverified.

Kobuko: Official data does not specify the server interval for continuous healing. The model's four ticks at 0.5-second intervals are a damage-conserving calculation convention.; The binary data provides castFrame=22 without a directly usable cast time. The model retains a provisional 0.3-second impact time.; Official text and binary data do not specify crit eligibility for the replacement magic attack, whether recasts refresh or stack charges, or when maximum Health is sampled. The model allows normal attack crits, refreshes to one charge and samples maximum Health when the attack launches.

9 units · 2 stars · 0 augments

Primal: turtle · Pixies 0

Transfigurist: the 25%/35%/50% bonus, AD/AP forms for five champions, Akali's bonus damage against burning enemies and recast on kill, Gromp's Purple Alpha Buff and Nidalee's third tracking spear are implemented. Equal same-scale bonuses use the roster's default form; official comparison/tie rules and unpublished geometry or coefficients remain unresolved.

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Primal: four selectable blessings, the two/four-unit thresholds, an augment's extra choice, Phoenix cross-combat takedown input and component eligibility are connected. Effects use the September 5, 2026 TFTips snapshot. Actual component delivery awaits the full-game economy flow.

Sivir: Sivir's official ability data is still a generic placeholder. Star-scaled coefficients and variable decomposition come from the same-day third-party pages/snapshot and remain provisional.; Current tft.tools 18.2 and the local snapshot list [3,3,6] extra bounces on kill; the same-day OP.GG page lists [3,6,10]. The conflict is retained pending official data rather than resolved from one source.; Nearby enemies has no published radius or selector. The 3-hex range, nearest-target rule, no consecutive hits on the same target and 0.15-second visual interval are provisional assumptions calibrated from video.; Official behavior when no other nearby enemy exists is unknown: stop, return to the caster or repeat the same target. The model stops the chain when there is no other valid target.

Nidalee: Coefficients come from TFTips and are not official ability variables.; AD-form leap selection and landing timing when changing targets within four hexes need calibration.

Vi: Vi's official Set 18 spell data is unavailable. The current values are third-party provisional data; the older TFTips snapshot is retained only as evidence of a version difference.; The video text says 6 seconds while current Lolchess says 3 seconds. The model follows the newer 3 seconds until official data resolves the version difference.; The video cannot verify all crowd-control immunity exceptions involving displacement, knockups or special forced controls. The model uses normal crowd-control immunity.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Akali: target-selection details when the AP form recasts

Ashe: Ashe's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The 8-hex length and 0.65-hex half-width are inferred from a board-piercing description, not official geometry. Exact collision width still needs official data.; The 20% Slow applies while enemies are inside the path. The engine approximates a continuous area with one-second ticks and a 0.05-second overlap refresh; actual server frequency is unknown.; The video confirms a piercing line and frost path but cannot independently establish tie-breaking when aiming at the most enemies. Direction follows the current ability text.

Master Yi: Official priority when AD and AP are equal

Yorick: CommunityDragon has character data for Yorick and his Spirit Walker, but the aggregate Set 18 snapshot contained only two shop champions. Active damage, healing and summon Health formulas still use provisional Lolchess values pending verification.; Spirit Walker data does not provide usable base AD, Armor or role classification. Unknown AD and Armor remain null in the source and use a neutral zero in combat; they are not inferred from the video's 27 damage.; The video confirms the order of the single-target active and the summon on death. Full summon attack rules, lifetime, inherited effects and retargeting when the target dies before impact remain unverified.

9 units · 2 stars · 0 augments

Primal: turtle · Pixies 0

Transfigurist: the 25%/35%/50% bonus, AD/AP forms for five champions, Akali's bonus damage against burning enemies and recast on kill, Gromp's Purple Alpha Buff and Nidalee's third tracking spear are implemented. Equal same-scale bonuses use the roster's default form; official comparison/tie rules and unpublished geometry or coefficients remain unresolved.

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Primal: four selectable blessings, the two/four-unit thresholds, an augment's extra choice, Phoenix cross-combat takedown input and component eligibility are connected. Effects use the September 5, 2026 TFTips snapshot. Actual component delivery awaits the full-game economy flow.

Vi: Vi's official Set 18 spell data is unavailable. The current values are third-party provisional data; the older TFTips snapshot is retained only as evidence of a version difference.; The video text says 6 seconds while current Lolchess says 3 seconds. The model follows the newer 3 seconds until official data resolves the version difference.; The video cannot verify all crowd-control immunity exceptions involving displacement, knockups or special forced controls. The model uses normal crowd-control immunity.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Sivir: Sivir's official ability data is still a generic placeholder. Star-scaled coefficients and variable decomposition come from the same-day third-party pages/snapshot and remain provisional.; Current tft.tools 18.2 and the local snapshot list [3,3,6] extra bounces on kill; the same-day OP.GG page lists [3,6,10]. The conflict is retained pending official data rather than resolved from one source.; Nearby enemies has no published radius or selector. The 3-hex range, nearest-target rule, no consecutive hits on the same target and 0.15-second visual interval are provisional assumptions calibrated from video.; Official behavior when no other nearby enemy exists is unknown: stop, return to the caster or repeat the same target. The model stops the chain when there is no other valid target.

Nidalee: Coefficients come from TFTips and are not official ability variables.; AD-form leap selection and landing timing when changing targets within four hexes need calibration.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Akali: target-selection details when the AP form recasts

Ashe: Ashe's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The 8-hex length and 0.65-hex half-width are inferred from a board-piercing description, not official geometry. Exact collision width still needs official data.; The 20% Slow applies while enemies are inside the path. The engine approximates a continuous area with one-second ticks and a 0.05-second overlap refresh; actual server frequency is unknown.; The video confirms a piercing line and frost path but cannot independently establish tie-breaking when aiming at the most enemies. Direction follows the current ability text.

Master Yi: Official priority when AD and AP are equal

Yorick: CommunityDragon has character data for Yorick and his Spirit Walker, but the aggregate Set 18 snapshot contained only two shop champions. Active damage, healing and summon Health formulas still use provisional Lolchess values pending verification.; Spirit Walker data does not provide usable base AD, Armor or role classification. Unknown AD and Armor remain null in the source and use a neutral zero in combat; they are not inferred from the video's 27 damage.; The video confirms the order of the single-target active and the summon on death. Full summon attack rules, lifetime, inherited effects and retargeting when the target dies before impact remain unverified.

7 units · 2 stars · 0 augments

Primal: — · Pixies 0

Solar: opening shields, shield and bonus magic-damage growth per unique 3-star ally, team Attack Speed/resistances at three 3-stars and conversion of 40% of bonus magic damage to true damage at five are implemented. Randomly upgrading a 3-star to 4-star every 4 seconds after reaching eight 3-stars is missing because 4-star ability data remains incomplete.

Kayle: Direction, area and maximum target count of the 3-star attack wave; Official ordering of Magic Resist reduction and bonus magic damage within one attack

Sejuani: Sejuani's official spell data is unavailable. Shields, the 10% maximum Health component, cone damage and line damage by star level use provisional values cross-checked between Lolchess and TFTips.; Video and text confirm a cone and a line but not precisely a 90-degree angle, 3-hex range, 0.5-hex half-width or 5-target cap. These remain provisional geometry pending official verification.; The delay between stages, reorientation after the target dies, shield-refresh merging and whether the line starts at the cone's end are unknown. The model uses 0.1 seconds, locks direction on cast, adds shields normally and uses the caster as the origin for both stages.

Leona: Leona's official Set 18 spell data is unavailable. Starting resistances, Armor scaling and stun values remain provisional, cross-checked between Lolchess and TFTips.; The video confirms the visible decay, target and stun but cannot establish the precise AP snapshot time. The model uses the usual combat-start snapshot.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Alune: Official data has not populated NumMoonshards, NumEnemies or MagicDamageCalc1/2.; Normal and full-moon coefficients by star level remain provisional.; The available direct binary payload conflicts with the current-set ability and cannot be promoted to verified data.

Fiddlesticks: The available binary data confirms only a 0.25-second cast and base champion data. The nearest three targets, 10 flat Magic Resist, 2 seconds, healing [410,485,850] and magic damage [70,105,170] remain provisional.; The video confirms three simultaneous visible tethers, periodic damage to multiple targets and periodic self-healing. It does not verify 10 Magic Resist, 2 seconds, star-scaled coefficients, server tick frequency or whether total healing changes with surviving targets.; The 0.5-second interval is a damage-conserving calculation convention, not a verified server frequency. One total self-healing pool per cast follows the third-party singular healing wording and still needs official verification.; Sources do not specify whether the drain locks normal attacks, can be interrupted by control, or ends individual tethers when targets leave range or die. The model schedules persistent events on locked targets without inventing extra channel-interruption rules.

6 units · 2 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Akali: target-selection details when the AP form recasts

Varus: Official ability variables are unavailable. The split of 385/580/925% AD plus 30/45/70% AP is consistent across two sources but remains provisional.; The projectile uses an inferred half-width of 0.65 hexes and a range of 10 hexes so the arrow can leave the board. Official text describes a straight arrow without specifying range or exact geometry.; Edited video cannot establish exact charge timing. Cast windup and recovery still use the engine's provisional defaults.

Shen: Shen's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Nearby has no official world-space radius. The model explicitly assumes 3 hexes pending official data or repeatable board tests.; Official tie-breaking among equally close injured allies and eligibility of minions or summons remain unverified. The engine currently excludes dummies and relics while allowing real champions.; The video verifies two shields and a subsequent empowered attack only. Current text says bonus magic damage, so the model adds it to the base attack rather than replacing that attack.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

7 units · 1 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Shen: Shen's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Nearby has no official world-space radius. The model explicitly assumes 3 hexes pending official data or repeatable board tests.; Official tie-breaking among equally close injured allies and eligibility of minions or summons remain unverified. The engine currently excludes dummies and relics while allowing real champions.; The video verifies two shields and a subsequent empowered attack only. Current text says bonus magic damage, so the model adds it to the base attack rather than replacing that attack.

Akali: target-selection details when the AP form recasts

Varus: Official ability variables are unavailable. The split of 385/580/925% AD plus 30/45/70% AP is consistent across two sources but remains provisional.; The projectile uses an inferred half-width of 0.65 hexes and a range of 10 hexes so the arrow can leave the board. Official text describes a straight arrow without specifying range or exact geometry.; Edited video cannot establish exact charge timing. Cast windup and recovery still use the engine's provisional defaults.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

9 units · 1 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Summoner: Yorick summon Health, Azir soldier damage and Raptor damage are connected. Zyra's +4/+6 attacks are included in the 12-second plant lifetime. Official attack intervals and spawn-selection rules are unavailable, so timing and positions remain provisional video-calibrated assumptions.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Ezreal: Ezreal's official spell values are unavailable. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Blink away has no exact published distance. The model allows up to 2 hexes and selects an available cell that maximizes distance from the nearest enemy, pending official or measurable video evidence.; The current English text does not say instead. The fourth cast therefore keeps its primary single-target damage and adds the wave. If official scripts show replacement behavior, the normal hit must be skipped.; The 8-hex range, 0.65-hex half-width and preference for the current target when hit counts tie are provisional geometry and tie-breaking, not official specifications.

Soraka: The supplied video does not show Soraka; footage of other champions is not used to infer her behavior.; Soraka's official Set 18 spell data is unavailable. Values, impact timing and exact visual spacing of the three extra stars still need official verification.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Azir: Azir's official Set 18 spell data is unavailable. The 150%, six attacks, two soldiers and current Lolchess values [40,60,96] remain provisional and conflict with older tft.tools values.; The older video caption says 6 seconds, while current Lolchess text says the next six attacks. The model follows the newer attack-count limit and retains the version difference in the audit.; Independent soldier crits, exact delay between damage packets, retargeting after the first packet kills and attack resets need official verification. Both packets currently lock the original target and share one normal-attack crit result.; The video does not establish soldiers as independent summons with Health, targetability or autonomous attacks. The model implements the explicitly described command damage only.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Yunara: The split now hits the nearest two enemies within 2 hexes of the primary target, replacing a 1-hex radius that often found only one. This follows the September 4, 2026 report; official text says nearby without a distance, so the radius remains provisional.; Yunara's official spell data is unavailable. Total primary damage [160,240,360] and the 35% split remain provisional. Based on Lolchess's AD/AP icons and variable magnitudes, the model maps ADDamage2 [150,225,335] to AD and ADDamage [10,15,25] to AP; official verification is still needed.; The video shows a short dash without a coordinate grid. The model moves one hex toward the locked target into an unoccupied cell. Official dash distance, passing through units and behavior with no valid forward cell are unknown.; The exact delay from the primary hit to the split is not visible. The model uses 0.1 seconds to preserve the hit-then-split order. The values divide evenly into [56,84,126], but no intermediate display rounding is applied pending confirmation of server rounding.

Yorick: CommunityDragon has character data for Yorick and his Spirit Walker, but the aggregate Set 18 snapshot contained only two shop champions. Active damage, healing and summon Health formulas still use provisional Lolchess values pending verification.; Spirit Walker data does not provide usable base AD, Armor or role classification. Unknown AD and Armor remain null in the source and use a neutral zero in combat; they are not inferred from the video's 27 damage.; The video confirms the order of the single-target active and the summon on death. Full summon attack rules, lifetime, inherited effects and retargeting when the target dies before impact remain unverified.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Ashe: Ashe's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The 8-hex length and 0.65-hex half-width are inferred from a board-piercing description, not official geometry. Exact collision width still needs official data.; The 20% Slow applies while enemies are inside the path. The engine approximates a continuous area with one-second ticks and a 0.05-second overlap refresh; actual server frequency is unknown.; The video confirms a piercing line and frost path but cannot independently establish tie-breaking when aiming at the most enemies. Direction follows the current ability text.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Summoner: Yorick summon Health, Azir soldier damage and Raptor damage are connected. Zyra's +4/+6 attacks are included in the 12-second plant lifetime. Official attack intervals and spawn-selection rules are unavailable, so timing and positions remain provisional video-calibrated assumptions.

Yunara: The split now hits the nearest two enemies within 2 hexes of the primary target, replacing a 1-hex radius that often found only one. This follows the September 4, 2026 report; official text says nearby without a distance, so the radius remains provisional.; Yunara's official spell data is unavailable. Total primary damage [160,240,360] and the 35% split remain provisional. Based on Lolchess's AD/AP icons and variable magnitudes, the model maps ADDamage2 [150,225,335] to AD and ADDamage [10,15,25] to AP; official verification is still needed.; The video shows a short dash without a coordinate grid. The model moves one hex toward the locked target into an unoccupied cell. Official dash distance, passing through units and behavior with no valid forward cell are unknown.; The exact delay from the primary hit to the split is not visible. The model uses 0.1 seconds to preserve the hit-then-split order. The values divide evenly into [56,84,126], but no intermediate display rounding is applied pending confirmation of server rounding.

Azir: Azir's official Set 18 spell data is unavailable. The 150%, six attacks, two soldiers and current Lolchess values [40,60,96] remain provisional and conflict with older tft.tools values.; The older video caption says 6 seconds, while current Lolchess text says the next six attacks. The model follows the newer attack-count limit and retains the version difference in the audit.; Independent soldier crits, exact delay between damage packets, retargeting after the first packet kills and attack resets need official verification. Both packets currently lock the original target and share one normal-attack crit result.; The video does not establish soldiers as independent summons with Health, targetability or autonomous attacks. The model implements the explicitly described command damage only.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Ezreal: Ezreal's official spell values are unavailable. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Blink away has no exact published distance. The model allows up to 2 hexes and selects an available cell that maximizes distance from the nearest enemy, pending official or measurable video evidence.; The current English text does not say instead. The fourth cast therefore keeps its primary single-target damage and adds the wave. If official scripts show replacement behavior, the normal hit must be skipped.; The 8-hex range, 0.65-hex half-width and preference for the current target when hit counts tie are provisional geometry and tie-breaking, not official specifications.

Soraka: The supplied video does not show Soraka; footage of other champions is not used to infer her behavior.; Soraka's official Set 18 spell data is unavailable. Values, impact timing and exact visual spacing of the three extra stars still need official verification.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Yorick: CommunityDragon has character data for Yorick and his Spirit Walker, but the aggregate Set 18 snapshot contained only two shop champions. Active damage, healing and summon Health formulas still use provisional Lolchess values pending verification.; Spirit Walker data does not provide usable base AD, Armor or role classification. Unknown AD and Armor remain null in the source and use a neutral zero in combat; they are not inferred from the video's 27 damage.; The video confirms the order of the single-target active and the summon on death. Full summon attack rules, lifetime, inherited effects and retargeting when the target dies before impact remain unverified.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Ashe: Ashe's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The 8-hex length and 0.65-hex half-width are inferred from a board-piercing description, not official geometry. Exact collision width still needs official data.; The 20% Slow applies while enemies are inside the path. The engine approximates a continuous area with one-second ticks and a 0.05-second overlap refresh; actual server frequency is unknown.; The video confirms a piercing line and frost path but cannot independently establish tie-breaking when aiming at the most enemies. Direction follows the current ability text.

8 units · 2 stars · 0 augments

Primal: — · Pixies 0

Rivals: With Ahead of the Curve active, the latest data confirms the link from Kha'Zix Mana restoration to Rengar at 50%, and from Rengar's actual self-healing to Kha'Zix at 100%. Persistent takedown counts for a lone Rival and the additional 3 progress for killing the other Rival are not connected.

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Elderwood: three separate combat entities, the total-star panel, death stun, periodic AD/AP, Deepwood damage and healing/Attack Speed on ally death are connected. Values use the September 5, 2026 TFTips snapshot. Additional rules at 11 units, secondary-hit geometry and healing scaling need official calibration.

Kha'Zix: Official ability data is unavailable. The two damage sets, 285/400/580 and 310/445/660, and their conditions agree across two sources but remain provisional.; The video cannot establish landing-cell selection or tie-breaking. The model selects the nearest empty adjacent cell to the caster, breaks ties consistently by row/column and uses a default 0.3-second cast pending official verification.; Official cancellation, tracking or in-place resolution is unknown when the target dies, becomes untargetable or has no valid empty adjacent cell before landing. The model cancels without redirecting to another target.

Rengar: Official spell data is unavailable. Damage and minimum/maximum healing use same-day provisional third-party values, with differences between older TFTips/tft.tools snapshots and newer sources.; The description says healing increases with the target's missing Health up to a maximum but provides no curve. The model linearly interpolates from minimum at 0% missing Health to maximum at 100%; this is not an official curve.; The video cannot establish the jump path, exact landing cell or windup/recovery. The model uses the nearest available cell and a 0.2-second hop.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Ezreal: Ezreal's official spell values are unavailable. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Blink away has no exact published distance. The model allows up to 2 hexes and selects an available cell that maximizes distance from the nearest enemy, pending official or measurable video evidence.; The current English text does not say instead. The fourth cast therefore keeps its primary single-target damage and adds the wave. If official scripts show replacement behavior, the normal hit must be skipped.; The 8-hex range, 0.65-hex half-width and preference for the current target when hit counts tie are provisional geometry and tie-breaking, not official specifications.

Gnar: Official leap selection, arc, collision half-width and exact landing timing lack binary data. The model uses legal landing cells within 3 hexes and approximates line collision from the thrown unit's origin to the farthest enemy.; Star-scaled coefficients use cross-checked provisional values and must not be labeled official exact values.

Alistar: Alistar's official spell data is unavailable. The three star-scaled value sets and 8% maximum Health component use provisional values cross-checked between Lolchess and TFTips.; The video caption says knockup, while current Lolchess text says a 1.5-second stun. The model follows the current text and does not label it a knockup; official verification is still required.; The third-party phrase all impairing effects has no official enumeration. The model clears external crowd control only, without also clearing Burn, Wound, Attack Speed reduction or Stasis. Retargeting if the current target dies before impact remains unverified.

8 units · 2 stars · 0 augments

Primal: — · Pixies 0

Rivals: With Ahead of the Curve active, the latest data confirms the link from Kha'Zix Mana restoration to Rengar at 50%, and from Rengar's actual self-healing to Kha'Zix at 100%. Persistent takedown counts for a lone Rival and the additional 3 progress for killing the other Rival are not connected.

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Elderwood: three separate combat entities, the total-star panel, death stun, periodic AD/AP, Deepwood damage and healing/Attack Speed on ally death are connected. Values use the September 5, 2026 TFTips snapshot. Additional rules at 11 units, secondary-hit geometry and healing scaling need official calibration.

Kha'Zix: Official ability data is unavailable. The two damage sets, 285/400/580 and 310/445/660, and their conditions agree across two sources but remain provisional.; The video cannot establish landing-cell selection or tie-breaking. The model selects the nearest empty adjacent cell to the caster, breaks ties consistently by row/column and uses a default 0.3-second cast pending official verification.; Official cancellation, tracking or in-place resolution is unknown when the target dies, becomes untargetable or has no valid empty adjacent cell before landing. The model cancels without redirecting to another target.

Rengar: Official spell data is unavailable. Damage and minimum/maximum healing use same-day provisional third-party values, with differences between older TFTips/tft.tools snapshots and newer sources.; The description says healing increases with the target's missing Health up to a maximum but provides no curve. The model linearly interpolates from minimum at 0% missing Health to maximum at 100%; this is not an official curve.; The video cannot establish the jump path, exact landing cell or windup/recovery. The model uses the nearest available cell and a 0.2-second hop.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Ezreal: Ezreal's official spell values are unavailable. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Blink away has no exact published distance. The model allows up to 2 hexes and selects an available cell that maximizes distance from the nearest enemy, pending official or measurable video evidence.; The current English text does not say instead. The fourth cast therefore keeps its primary single-target damage and adds the wave. If official scripts show replacement behavior, the normal hit must be skipped.; The 8-hex range, 0.65-hex half-width and preference for the current target when hit counts tie are provisional geometry and tie-breaking, not official specifications.

Gnar: Official leap selection, arc, collision half-width and exact landing timing lack binary data. The model uses legal landing cells within 3 hexes and approximates line collision from the thrown unit's origin to the farthest enemy.; Star-scaled coefficients use cross-checked provisional values and must not be labeled official exact values.

Alistar: Alistar's official spell data is unavailable. The three star-scaled value sets and 8% maximum Health component use provisional values cross-checked between Lolchess and TFTips.; The video caption says knockup, while current Lolchess text says a 1.5-second stun. The model follows the current text and does not label it a knockup; official verification is still required.; The third-party phrase all impairing effects has no official enumeration. The model clears external crowd control only, without also clearing Burn, Wound, Attack Speed reduction or Stasis. Retargeting if the current target dies before impact remains unverified.

7 units · 2 stars · 0 augments

Primal: — · Pixies 0

Solar: opening shields, shield and bonus magic-damage growth per unique 3-star ally, team Attack Speed/resistances at three 3-stars and conversion of 40% of bonus magic damage to true damage at five are implemented. Randomly upgrading a 3-star to 4-star every 4 seconds after reaching eight 3-stars is missing because 4-star ability data remains incomplete.

Leona: Leona's official Set 18 spell data is unavailable. Starting resistances, Armor scaling and stun values remain provisional, cross-checked between Lolchess and TFTips.; The video confirms the visible decay, target and stun but cannot establish the precise AP snapshot time. The model uses the usual combat-start snapshot.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Kayle: Direction, area and maximum target count of the 3-star attack wave; Official ordering of Magic Resist reduction and bonus magic damage within one attack

Sejuani: Sejuani's official spell data is unavailable. Shields, the 10% maximum Health component, cone damage and line damage by star level use provisional values cross-checked between Lolchess and TFTips.; Video and text confirm a cone and a line but not precisely a 90-degree angle, 3-hex range, 0.5-hex half-width or 5-target cap. These remain provisional geometry pending official verification.; The delay between stages, reorientation after the target dies, shield-refresh merging and whether the line starts at the cone's end are unknown. The model uses 0.1 seconds, locks direction on cast, adds shields normally and uses the caster as the origin for both stages.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Alune: Official data has not populated NumMoonshards, NumEnemies or MagicDamageCalc1/2.; Normal and full-moon coefficients by star level remain provisional.; The available direct binary payload conflicts with the current-set ability and cannot be promoted to verified data.

Fiddlesticks: The available binary data confirms only a 0.25-second cast and base champion data. The nearest three targets, 10 flat Magic Resist, 2 seconds, healing [410,485,850] and magic damage [70,105,170] remain provisional.; The video confirms three simultaneous visible tethers, periodic damage to multiple targets and periodic self-healing. It does not verify 10 Magic Resist, 2 seconds, star-scaled coefficients, server tick frequency or whether total healing changes with surviving targets.; The 0.5-second interval is a damage-conserving calculation convention, not a verified server frequency. One total self-healing pool per cast follows the third-party singular healing wording and still needs official verification.; Sources do not specify whether the drain locks normal attacks, can be interrupted by control, or ends individual tethers when targets leave range or die. The model schedules persistent events on locked targets without inventing extra channel-interruption rules.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Fae: with an explicit Pixie count, each Pixie grants Fae units 5%/8% AD and AP, plus a one-time heal below 50% Health for 2.5%/4% maximum Health per Pixie. Little Buddy provides attraction thresholds and Golden Pixie rewards. Automatic damage/heal/shield accumulation and cross-combat gold settlement are not connected.

Cinderling: The five leaves now hit sequentially within 0.5 seconds instead of simultaneously, following the September 4, 2026 report that they should deal five hits. The official interval between leaves is unknown.; Official ability variables are unavailable. The values 340/510/765% AD plus 30/45/70% AP, 4 seconds, 20% and 1% agree across two sources but remain provisional.; The video confirms five projectiles but not independent crit rolls or on-hit effects for each. The model processes five sequential damage events within 0.5 seconds.; Crimson Buff's condition and external state for granting the Burning Sapling 22% AD per cast are not connected, so the bonus is not automatically applied to Bough.

Caitlyn: Caitlyn's official spell data is unavailable. Two sources agree on every third attack and the split of AD [200,300,500] plus AP [20,30,45], but these values remain provisional.; The video caption could mean the attack after completing three attacks. The video supports a periodic replacement attack; the exact every-third-attack rule follows the current Lolchess and TFTips text.; Official projectile travel, crit eligibility and on-hit ordering are undocumented. The model uses normal attack resolution, allowing normal crits and triggering attack items, traits and on-hit effects.

Tristana: Official Set 18 spell values are unavailable for field-by-field verification. Current values use TFTips as a provisional source.

Sivir: Sivir's official ability data is still a generic placeholder. Star-scaled coefficients and variable decomposition come from the same-day third-party pages/snapshot and remain provisional.; Current tft.tools 18.2 and the local snapshot list [3,3,6] extra bounces on kill; the same-day OP.GG page lists [3,6,10]. The conflict is retained pending official data rather than resolved from one source.; Nearby enemies has no published radius or selector. The 3-hex range, nearest-target rule, no consecutive hits on the same target and 0.15-second visual interval are provisional assumptions calibrated from video.; Official behavior when no other nearby enemy exists is unknown: stop, return to the caster or repeat the same target. The model stops the chain when there is no other valid target.

Ashe: Ashe's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The 8-hex length and 0.65-hex half-width are inferred from a board-piercing description, not official geometry. Exact collision width still needs official data.; The 20% Slow applies while enemies are inside the path. The engine approximates a continuous area with one-second ticks and a 0.05-second overlap refresh; actual server frequency is unknown.; The video confirms a piercing line and frost path but cannot independently establish tie-breaking when aiming at the most enemies. Direction follows the current ability text.

Lillia: Lillia's official ability data is unavailable. Healing [325,475,800], magic damage [90,135,2000], sleep [1.5,1.75,8], four butterflies, the 1,000-damage threshold and 10% maximum Health bonus damage agree across two sources but remain provisional.; The captions at 266–269 seconds show an older five-target / 2-second version that conflicts with current sources. The video supports visible behavior, not the current target count or duration.; Sources do not specify whether the initial hit counts toward 1,000, whether the threshold is before/after mitigation or shields, whether natural waking adds damage, or whether butterflies can repeat targets. The model counts the initial hit after mitigation but before shields, adds no damage on natural expiry and chooses the nearest four unique enemies.

Rammus: Official taunt radius and duration; the current assumptions are 2 hexes and 4 seconds; Official calculation of shield-break damage with ADDamageRatio 0.5

Morgana: Morgana's official ability values are unavailable. Only the same-day Lolchess page fully lists the three value sets, 4 seconds, nearest three enemies and 25% Omnivamp, so these remain provisional.; Whether the area stays fixed or follows the caster is unverified. It currently stays at the cast location; following behavior is not claimed as verified.; Ticks occur on complete second boundaries from impact +1 through impact +4. The 0.3-second windup, first-tick timing and ordering with expiry still need official scripts.

Sentinel: Ancient Sentinel's official spell values are unavailable. Shield, maximum Health scaling, damage and 10 Mana Reave use provisional values from the current tft.tools/Lolchess snapshot.; The ability says a brief knockup without a duration. The engine does not invent 0.1, 0.25 or another duration that would change combat scheduling.; The fissure uses an inferred 8-hex length and 0.5-hex half-width based on the board-wide description. Text supports aiming at the largest group, but exact collision width requires official data or direct video evidence.; Blue Buff's 2 Mana Regen for allies per cast applies only when Sentinel receives the matching Alpha/Blue bonus. That selection state is not connected to the unit, so it is excluded from the base ability.

9 units · 1 stars · 0 augments

Primal: — · Pixies 0

Fae: with an explicit Pixie count, each Pixie grants Fae units 5%/8% AD and AP, plus a one-time heal below 50% Health for 2.5%/4% maximum Health per Pixie. Little Buddy provides attraction thresholds and Golden Pixie rewards. Automatic damage/heal/shield accumulation and cross-combat gold settlement are not connected.

Ashe: Ashe's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The 8-hex length and 0.65-hex half-width are inferred from a board-piercing description, not official geometry. Exact collision width still needs official data.; The 20% Slow applies while enemies are inside the path. The engine approximates a continuous area with one-second ticks and a 0.05-second overlap refresh; actual server frequency is unknown.; The video confirms a piercing line and frost path but cannot independently establish tie-breaking when aiming at the most enemies. Direction follows the current ability text.

Sivir: Sivir's official ability data is still a generic placeholder. Star-scaled coefficients and variable decomposition come from the same-day third-party pages/snapshot and remain provisional.; Current tft.tools 18.2 and the local snapshot list [3,3,6] extra bounces on kill; the same-day OP.GG page lists [3,6,10]. The conflict is retained pending official data rather than resolved from one source.; Nearby enemies has no published radius or selector. The 3-hex range, nearest-target rule, no consecutive hits on the same target and 0.15-second visual interval are provisional assumptions calibrated from video.; Official behavior when no other nearby enemy exists is unknown: stop, return to the caster or repeat the same target. The model stops the chain when there is no other valid target.

Tristana: Official Set 18 spell values are unavailable for field-by-field verification. Current values use TFTips as a provisional source.

Caitlyn: Caitlyn's official spell data is unavailable. Two sources agree on every third attack and the split of AD [200,300,500] plus AP [20,30,45], but these values remain provisional.; The video caption could mean the attack after completing three attacks. The video supports a periodic replacement attack; the exact every-third-attack rule follows the current Lolchess and TFTips text.; Official projectile travel, crit eligibility and on-hit ordering are undocumented. The model uses normal attack resolution, allowing normal crits and triggering attack items, traits and on-hit effects.

Cinderling: The five leaves now hit sequentially within 0.5 seconds instead of simultaneously, following the September 4, 2026 report that they should deal five hits. The official interval between leaves is unknown.; Official ability variables are unavailable. The values 340/510/765% AD plus 30/45/70% AP, 4 seconds, 20% and 1% agree across two sources but remain provisional.; The video confirms five projectiles but not independent crit rolls or on-hit effects for each. The model processes five sequential damage events within 0.5 seconds.; Crimson Buff's condition and external state for granting the Burning Sapling 22% AD per cast are not connected, so the bonus is not automatically applied to Bough.

Lillia: Lillia's official ability data is unavailable. Healing [325,475,800], magic damage [90,135,2000], sleep [1.5,1.75,8], four butterflies, the 1,000-damage threshold and 10% maximum Health bonus damage agree across two sources but remain provisional.; The captions at 266–269 seconds show an older five-target / 2-second version that conflicts with current sources. The video supports visible behavior, not the current target count or duration.; Sources do not specify whether the initial hit counts toward 1,000, whether the threshold is before/after mitigation or shields, whether natural waking adds damage, or whether butterflies can repeat targets. The model counts the initial hit after mitigation but before shields, adds no damage on natural expiry and chooses the nearest four unique enemies.

Rammus: Official taunt radius and duration; the current assumptions are 2 hexes and 4 seconds; Official calculation of shield-break damage with ADDamageRatio 0.5

Morgana: Morgana's official ability values are unavailable. Only the same-day Lolchess page fully lists the three value sets, 4 seconds, nearest three enemies and 25% Omnivamp, so these remain provisional.; Whether the area stays fixed or follows the caster is unverified. It currently stays at the cast location; following behavior is not claimed as verified.; Ticks occur on complete second boundaries from impact +1 through impact +4. The 0.3-second windup, first-tick timing and ordering with expiry still need official scripts.

Sentinel: Ancient Sentinel's official spell values are unavailable. Shield, maximum Health scaling, damage and 10 Mana Reave use provisional values from the current tft.tools/Lolchess snapshot.; The ability says a brief knockup without a duration. The engine does not invent 0.1, 0.25 or another duration that would change combat scheduling.; The fissure uses an inferred 8-hex length and 0.5-hex half-width based on the board-wide description. Text supports aiming at the largest group, but exact collision width requires official data or direct video evidence.; Blue Buff's 2 Mana Regen for allies per cast applies only when Sentinel receives the matching Alpha/Blue bonus. That selection state is not connected to the unit, so it is excluded from the base ability.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Coven: single-combat Essence settlement grants 2/2/3/10 Essence per kill at 3/4/5/7 Coven, plus 18/25/32/60 for a loss, and reports the first 40-Essence ritual threshold. Persistence, increasing thresholds and the full reward table remain incomplete.

Despoiler: damage resolution is fully connected in combat. The older one-sided base-DPS preview lacks an enemy Health pool and can apply only the base amplification.

Morgana: Morgana's official ability values are unavailable. Only the same-day Lolchess page fully lists the three value sets, 4 seconds, nearest three enemies and 25% Omnivamp, so these remain provisional.; Whether the area stays fixed or follows the caster is unverified. It currently stays at the cast location; following behavior is not claimed as verified.; Ticks occur on complete second boundaries from impact +1 through impact +4. The 0.3-second windup, first-tick timing and ordering with expiry still need official scripts.

Cassiopeia: Official ability data is unavailable. The values 440/660/1050, 15 seconds, two targets and stacking agree across two sources but remain provisional and require official verification.; Neither the video nor third-party text reveals the server's damage-tick interval. One-second ticks are a damage-conserving calculation convention, not a verified server frequency.; It is unknown whether nearest is measured from the caster or primary target, whether two Cassiopeias share the unpoisoned-target check, and whether they retarget after a death before impact. The model measures from the primary target, shares poison-family marks and cancels when that target becomes invalid.

Caitlyn: Caitlyn's official spell data is unavailable. Two sources agree on every third attack and the split of AD [200,300,500] plus AP [20,30,45], but these values remain provisional.; The video caption could mean the attack after completing three attacks. The video supports a periodic replacement attack; the exact every-third-attack rule follows the current Lolchess and TFTips text.; Official projectile travel, crit eligibility and on-hit ordering are undocumented. The model uses normal attack resolution, allowing normal crits and triggering attack items, traits and on-hit effects.

Elise: Complete official spell data is still unavailable; the values remain provisional.; The transformation animation does not itself change numerical results. Impact timing must be recalibrated if official data establishes a different animation duration.

Camille: Camille's official spell data is unavailable. Two sources agree on physical damage with AD values [160,240,410], AP values [10,15,25], and shields [60,90,200]. The model applies the AD and AP scaling separately; the values remain provisional.; The video caption at 23 seconds lists a 3-second shield; the newer Lolchess description lists 2 seconds. The model uses 2 seconds and treats the video as evidence of the slash and shield animation only.; Retargeting after a target dies before impact, exact cast windup, shield modifiers and independent expiry of multiple shields are undocumented. The model retains target lock, a default 0.3-second cast and a shared shield expiry time.

Sentinel: Ancient Sentinel's official spell values are unavailable. Shield, maximum Health scaling, damage and 10 Mana Reave use provisional values from the current tft.tools/Lolchess snapshot.; The ability says a brief knockup without a duration. The engine does not invent 0.1, 0.25 or another duration that would change combat scheduling.; The fissure uses an inferred 8-hex length and 0.5-hex half-width based on the board-wide description. Text supports aiming at the largest group, but exact collision width requires official data or direct video evidence.; Blue Buff's 2 Mana Regen for allies per cast applies only when Sentinel receives the matching Alpha/Blue bonus. That selection state is not connected to the unit, so it is excluded from the base ability.

Brambleback: Red Buff variant: attacks apply 1% Burn and heal 4% maximum Health; Official leap landing rule; the model uses the nearest empty cell beside the target

Diana: Diana's official spell data is unavailable. Six orbs, 2 hexes, damage per orb [65,100,155], shields [150,225,300] and 2 seconds agree across two sources but remain provisional.; Sources do not specify whether six orbs are distributed evenly, randomly or by other weights among valid enemies. The model assigns each orb independently and records the decisions; this is not an official selection rule.; A clear current-version video, precise projectile travel, behavior after death before impact and independent expiry of multiple shields are unverified. The model uses a default 0.3-second impact delay and shared shield expiry.

Alune: Official data has not populated NumMoonshards, NumEnemies or MagicDamageCalc1/2.; Normal and full-moon coefficients by star level remain provisional.; The available direct binary payload conflicts with the current-set ability and cannot be promoted to verified data.

8 units · 2 stars · 0 augments

Primal: turtle · Pixies 0

Transfigurist: the 25%/35%/50% bonus, AD/AP forms for five champions, Akali's bonus damage against burning enemies and recast on kill, Gromp's Purple Alpha Buff and Nidalee's third tracking spear are implemented. Equal same-scale bonuses use the roster's default form; official comparison/tie rules and unpublished geometry or coefficients remain unresolved.

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Primal: four selectable blessings, the two/four-unit thresholds, an augment's extra choice, Phoenix cross-combat takedown input and component eligibility are connected. Effects use the September 5, 2026 TFTips snapshot. Actual component delivery awaits the full-game economy flow.

Sivir: Sivir's official ability data is still a generic placeholder. Star-scaled coefficients and variable decomposition come from the same-day third-party pages/snapshot and remain provisional.; Current tft.tools 18.2 and the local snapshot list [3,3,6] extra bounces on kill; the same-day OP.GG page lists [3,6,10]. The conflict is retained pending official data rather than resolved from one source.; Nearby enemies has no published radius or selector. The 3-hex range, nearest-target rule, no consecutive hits on the same target and 0.15-second visual interval are provisional assumptions calibrated from video.; Official behavior when no other nearby enemy exists is unknown: stop, return to the caster or repeat the same target. The model stops the chain when there is no other valid target.

Nidalee: Coefficients come from TFTips and are not official ability variables.; AD-form leap selection and landing timing when changing targets within four hexes need calibration.

Vi: Vi's official Set 18 spell data is unavailable. The current values are third-party provisional data; the older TFTips snapshot is retained only as evidence of a version difference.; The video text says 6 seconds while current Lolchess says 3 seconds. The model follows the newer 3 seconds until official data resolves the version difference.; The video cannot verify all crowd-control immunity exceptions involving displacement, knockups or special forced controls. The model uses normal crowd-control immunity.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Akali: target-selection details when the AP form recasts

Ashe: Ashe's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The 8-hex length and 0.65-hex half-width are inferred from a board-piercing description, not official geometry. Exact collision width still needs official data.; The 20% Slow applies while enemies are inside the path. The engine approximates a continuous area with one-second ticks and a 0.05-second overlap refresh; actual server frequency is unknown.; The video confirms a piercing line and frost path but cannot independently establish tie-breaking when aiming at the most enemies. Direction follows the current ability text.

Master Yi: Official priority when AD and AP are equal

8 units · 2 stars · 0 augments

Primal: turtle · Pixies 0

Transfigurist: the 25%/35%/50% bonus, AD/AP forms for five champions, Akali's bonus damage against burning enemies and recast on kill, Gromp's Purple Alpha Buff and Nidalee's third tracking spear are implemented. Equal same-scale bonuses use the roster's default form; official comparison/tie rules and unpublished geometry or coefficients remain unresolved.

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Primal: four selectable blessings, the two/four-unit thresholds, an augment's extra choice, Phoenix cross-combat takedown input and component eligibility are connected. Effects use the September 5, 2026 TFTips snapshot. Actual component delivery awaits the full-game economy flow.

Vi: Vi's official Set 18 spell data is unavailable. The current values are third-party provisional data; the older TFTips snapshot is retained only as evidence of a version difference.; The video text says 6 seconds while current Lolchess says 3 seconds. The model follows the newer 3 seconds until official data resolves the version difference.; The video cannot verify all crowd-control immunity exceptions involving displacement, knockups or special forced controls. The model uses normal crowd-control immunity.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Sivir: Sivir's official ability data is still a generic placeholder. Star-scaled coefficients and variable decomposition come from the same-day third-party pages/snapshot and remain provisional.; Current tft.tools 18.2 and the local snapshot list [3,3,6] extra bounces on kill; the same-day OP.GG page lists [3,6,10]. The conflict is retained pending official data rather than resolved from one source.; Nearby enemies has no published radius or selector. The 3-hex range, nearest-target rule, no consecutive hits on the same target and 0.15-second visual interval are provisional assumptions calibrated from video.; Official behavior when no other nearby enemy exists is unknown: stop, return to the caster or repeat the same target. The model stops the chain when there is no other valid target.

Nidalee: Coefficients come from TFTips and are not official ability variables.; AD-form leap selection and landing timing when changing targets within four hexes need calibration.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Akali: target-selection details when the AP form recasts

Ashe: Ashe's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The 8-hex length and 0.65-hex half-width are inferred from a board-piercing description, not official geometry. Exact collision width still needs official data.; The 20% Slow applies while enemies are inside the path. The engine approximates a continuous area with one-second ticks and a 0.05-second overlap refresh; actual server frequency is unknown.; The video confirms a piercing line and frost path but cannot independently establish tie-breaking when aiming at the most enemies. Direction follows the current ability text.

Master Yi: Official priority when AD and AP are equal

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Coven: single-combat Essence settlement grants 2/2/3/10 Essence per kill at 3/4/5/7 Coven, plus 18/25/32/60 for a loss, and reports the first 40-Essence ritual threshold. Persistence, increasing thresholds and the full reward table remain incomplete.

Despoiler: damage resolution is fully connected in combat. The older one-sided base-DPS preview lacks an enemy Health pool and can apply only the base amplification.

Camille: Camille's official spell data is unavailable. Two sources agree on physical damage with AD values [160,240,410], AP values [10,15,25], and shields [60,90,200]. The model applies the AD and AP scaling separately; the values remain provisional.; The video caption at 23 seconds lists a 3-second shield; the newer Lolchess description lists 2 seconds. The model uses 2 seconds and treats the video as evidence of the slash and shield animation only.; Retargeting after a target dies before impact, exact cast windup, shield modifiers and independent expiry of multiple shields are undocumented. The model retains target lock, a default 0.3-second cast and a shared shield expiry time.

Caitlyn: Caitlyn's official spell data is unavailable. Two sources agree on every third attack and the split of AD [200,300,500] plus AP [20,30,45], but these values remain provisional.; The video caption could mean the attack after completing three attacks. The video supports a periodic replacement attack; the exact every-third-attack rule follows the current Lolchess and TFTips text.; Official projectile travel, crit eligibility and on-hit ordering are undocumented. The model uses normal attack resolution, allowing normal crits and triggering attack items, traits and on-hit effects.

Elise: Complete official spell data is still unavailable; the values remain provisional.; The transformation animation does not itself change numerical results. Impact timing must be recalibrated if official data establishes a different animation duration.

Cassiopeia: Official ability data is unavailable. The values 440/660/1050, 15 seconds, two targets and stacking agree across two sources but remain provisional and require official verification.; Neither the video nor third-party text reveals the server's damage-tick interval. One-second ticks are a damage-conserving calculation convention, not a verified server frequency.; It is unknown whether nearest is measured from the caster or primary target, whether two Cassiopeias share the unpoisoned-target check, and whether they retarget after a death before impact. The model measures from the primary target, shares poison-family marks and cancels when that target becomes invalid.

Morgana: Morgana's official ability values are unavailable. Only the same-day Lolchess page fully lists the three value sets, 4 seconds, nearest three enemies and 25% Omnivamp, so these remain provisional.; Whether the area stays fixed or follows the caster is unverified. It currently stays at the cast location; following behavior is not claimed as verified.; Ticks occur on complete second boundaries from impact +1 through impact +4. The 0.3-second windup, first-tick timing and ordering with expiry still need official scripts.

Sentinel: Ancient Sentinel's official spell values are unavailable. Shield, maximum Health scaling, damage and 10 Mana Reave use provisional values from the current tft.tools/Lolchess snapshot.; The ability says a brief knockup without a duration. The engine does not invent 0.1, 0.25 or another duration that would change combat scheduling.; The fissure uses an inferred 8-hex length and 0.5-hex half-width based on the board-wide description. Text supports aiming at the largest group, but exact collision width requires official data or direct video evidence.; Blue Buff's 2 Mana Regen for allies per cast applies only when Sentinel receives the matching Alpha/Blue bonus. That selection state is not connected to the unit, so it is excluded from the base ability.

Brambleback: Red Buff variant: attacks apply 1% Burn and heal 4% maximum Health; Official leap landing rule; the model uses the nearest empty cell beside the target

Diana: Diana's official spell data is unavailable. Six orbs, 2 hexes, damage per orb [65,100,155], shields [150,225,300] and 2 seconds agree across two sources but remain provisional.; Sources do not specify whether six orbs are distributed evenly, randomly or by other weights among valid enemies. The model assigns each orb independently and records the decisions; this is not an official selection rule.; A clear current-version video, precise projectile travel, behavior after death before impact and independent expiry of multiple shields are unverified. The model uses a default 0.3-second impact delay and shared shield expiry.

Alune: Official data has not populated NumMoonshards, NumEnemies or MagicDamageCalc1/2.; Normal and full-moon coefficients by star level remain provisional.; The available direct binary payload conflicts with the current-set ability and cannot be promoted to verified data.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Fae: with an explicit Pixie count, each Pixie grants Fae units 5%/8% AD and AP, plus a one-time heal below 50% Health for 2.5%/4% maximum Health per Pixie. Little Buddy provides attraction thresholds and Golden Pixie rewards. Automatic damage/heal/shield accumulation and cross-combat gold settlement are not connected.

Varus: Official ability variables are unavailable. The split of 385/580/925% AD plus 30/45/70% AP is consistent across two sources but remains provisional.; The projectile uses an inferred half-width of 0.65 hexes and a range of 10 hexes so the arrow can leave the board. Official text describes a straight arrow without specifying range or exact geometry.; Edited video cannot establish exact charge timing. Cast windup and recovery still use the engine's provisional defaults.

Xayah: Xayah's official Set 18 spell data is unavailable. The values 72/108/165, 50%, 5 attacks and 2 Armor reduction per attack remain provisional, cross-checked against third-party text and video captions.; The video shows feather paths but does not establish piercing or damage to secondary targets. The model follows the text and hits the current attack target without adding unsupported area damage.; Official data does not yet specify attack resets, missile speed or exact travel time.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Kayle: Direction, area and maximum target count of the 3-star attack wave; Official ordering of Magic Resist reduction and bonus magic damage within one attack

Mama Beak: Official character data identifies the mother and baby Raptors, but the ability is still a generic placeholder. Four birds, 5 seconds and damage [27,41,65] agree across two third-party pages but remain provisional.; The video confirms separate visible models and coordinated attacks, but not spawn cells, movement, travel time or collision. The model uses synchronized damage proxies that do not occupy board cells.; Summon crits, on-hit/item triggers, hits after the mother's attack kills the target, and stacking versus refreshing on recast lack official evidence. The model currently uses no crits, no triggers, no retargeting and refreshes.; The 0.25-second cast uses the shared spell placeholder's timing. It is not a calibrated official windup.; The source and stacking rules of the orange buff are unpublished. Armor reduction is applied only when that external effect is selected.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Lillia: Lillia's official ability data is unavailable. Healing [325,475,800], magic damage [90,135,2000], sleep [1.5,1.75,8], four butterflies, the 1,000-damage threshold and 10% maximum Health bonus damage agree across two sources but remain provisional.; The captions at 266–269 seconds show an older five-target / 2-second version that conflicts with current sources. The video supports visible behavior, not the current target count or duration.; Sources do not specify whether the initial hit counts toward 1,000, whether the threshold is before/after mitigation or shields, whether natural waking adds damage, or whether butterflies can repeat targets. The model counts the initial hit after mitigation but before shields, adds no damage on natural expiry and chooses the nearest four unique enemies.

Shen: Shen's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Nearby has no official world-space radius. The model explicitly assumes 3 hexes pending official data or repeatable board tests.; Official tie-breaking among equally close injured allies and eligibility of minions or summons remain unverified. The engine currently excludes dummies and relics while allowing real champions.; The video verifies two shields and a subsequent empowered attack only. Current text says bonus magic damage, so the model adds it to the base attack rather than replacing that attack.

Leona: Leona's official Set 18 spell data is unavailable. Starting resistances, Armor scaling and stun values remain provisional, cross-checked between Lolchess and TFTips.; The video confirms the visible decay, target and stun but cannot establish the precise AP snapshot time. The model uses the usual combat-start snapshot.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Ornn: The cone now has a 1-hex range, 120-degree angle and maximum of 3 targets, following the September 4, 2026 screenshot report that the breath mainly hits the unit directly ahead. Official text gives no range, so the geometry remains provisional.; Official ability variables are unavailable. Two sources agree on 400/460/550 shields and 200/300/450 damage, but the values remain provisional.; The video confirms a short-range cone but cannot establish a 90-degree angle, 3-hex range or 5-target limit precisely. These remain provisional geometry.; The Artifact Anvil reward at 90,000 accumulated damage taken, with doubled progress at 3 stars, is a cross-combat economy quest and is not connected to the combat engine.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Fae: with an explicit Pixie count, each Pixie grants Fae units 5%/8% AD and AP, plus a one-time heal below 50% Health for 2.5%/4% maximum Health per Pixie. Little Buddy provides attraction thresholds and Golden Pixie rewards. Automatic damage/heal/shield accumulation and cross-combat gold settlement are not connected.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Mama Beak: Official character data identifies the mother and baby Raptors, but the ability is still a generic placeholder. Four birds, 5 seconds and damage [27,41,65] agree across two third-party pages but remain provisional.; The video confirms separate visible models and coordinated attacks, but not spawn cells, movement, travel time or collision. The model uses synchronized damage proxies that do not occupy board cells.; Summon crits, on-hit/item triggers, hits after the mother's attack kills the target, and stacking versus refreshing on recast lack official evidence. The model currently uses no crits, no triggers, no retargeting and refreshes.; The 0.25-second cast uses the shared spell placeholder's timing. It is not a calibrated official windup.; The source and stacking rules of the orange buff are unpublished. Armor reduction is applied only when that external effect is selected.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Kayle: Direction, area and maximum target count of the 3-star attack wave; Official ordering of Magic Resist reduction and bonus magic damage within one attack

Varus: Official ability variables are unavailable. The split of 385/580/925% AD plus 30/45/70% AP is consistent across two sources but remains provisional.; The projectile uses an inferred half-width of 0.65 hexes and a range of 10 hexes so the arrow can leave the board. Official text describes a straight arrow without specifying range or exact geometry.; Edited video cannot establish exact charge timing. Cast windup and recovery still use the engine's provisional defaults.

Xayah: Xayah's official Set 18 spell data is unavailable. The values 72/108/165, 50%, 5 attacks and 2 Armor reduction per attack remain provisional, cross-checked against third-party text and video captions.; The video shows feather paths but does not establish piercing or damage to secondary targets. The model follows the text and hits the current attack target without adding unsupported area damage.; Official data does not yet specify attack resets, missile speed or exact travel time.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Lillia: Lillia's official ability data is unavailable. Healing [325,475,800], magic damage [90,135,2000], sleep [1.5,1.75,8], four butterflies, the 1,000-damage threshold and 10% maximum Health bonus damage agree across two sources but remain provisional.; The captions at 266–269 seconds show an older five-target / 2-second version that conflicts with current sources. The video supports visible behavior, not the current target count or duration.; Sources do not specify whether the initial hit counts toward 1,000, whether the threshold is before/after mitigation or shields, whether natural waking adds damage, or whether butterflies can repeat targets. The model counts the initial hit after mitigation but before shields, adds no damage on natural expiry and chooses the nearest four unique enemies.

Shen: Shen's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Nearby has no official world-space radius. The model explicitly assumes 3 hexes pending official data or repeatable board tests.; Official tie-breaking among equally close injured allies and eligibility of minions or summons remain unverified. The engine currently excludes dummies and relics while allowing real champions.; The video verifies two shields and a subsequent empowered attack only. Current text says bonus magic damage, so the model adds it to the base attack rather than replacing that attack.

Leona: Leona's official Set 18 spell data is unavailable. Starting resistances, Armor scaling and stun values remain provisional, cross-checked between Lolchess and TFTips.; The video confirms the visible decay, target and stun but cannot establish the precise AP snapshot time. The model uses the usual combat-start snapshot.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Ornn: The cone now has a 1-hex range, 120-degree angle and maximum of 3 targets, following the September 4, 2026 screenshot report that the breath mainly hits the unit directly ahead. Official text gives no range, so the geometry remains provisional.; Official ability variables are unavailable. Two sources agree on 400/460/550 shields and 200/300/450 damage, but the values remain provisional.; The video confirms a short-range cone but cannot establish a 90-degree angle, 3-hex range or 5-target limit precisely. These remain provisional geometry.; The Artifact Anvil reward at 90,000 accumulated damage taken, with doubled progress at 3 stars, is a cross-combat economy quest and is not connected to the combat engine.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Veteran: explicit rider selection grants Health/Attack Speed of 15%/15%, 40%/35% or 45%/45%. Little Buddy provides melee healing, ranged bonuses and buddy effects. Sharing buddy abilities at 50%/100% at five/seven units is not implemented, and buddy cast timing still needs verification.

Elderwood: three separate combat entities, the total-star panel, death stun, periodic AD/AP, Deepwood damage and healing/Attack Speed on ally death are connected. Values use the September 5, 2026 TFTips snapshot. Additional rules at 11 units, secondary-hit geometry and healing scaling need official calibration.

Blackthorn: explicit cell selection, pre-combat sacrifice at 2/4/6 Blackthorn and +175/+350/+600 flat team Health are implemented. Little Buddy supplies sacrifice tables by role, star level, cost and breakpoint. Member bonuses are not implemented; rounding and high-star Mana Regen values need another cross-check.

Kobuko: Official data does not specify the server interval for continuous healing. The model's four ticks at 0.5-second intervals are a damage-conserving calculation convention.; The binary data provides castFrame=22 without a directly usable cast time. The model retains a provisional 0.3-second impact time.; Official text and binary data do not specify crit eligibility for the replacement magic attack, whether recasts refresh or stack charges, or when maximum Health is sampled. The model allows normal attack crits, refreshes to one charge and samples maximum Health when the attack launches.

Rek'Sai: Rek'Sai's official spell data is unavailable. Healing [10,14,20], the 1% maximum Health component, magic damage [70,105,160] and timing agree across two sources but remain provisional.; The video describes an older version with triple healing for 5 seconds after casting; current Lolchess text says 3 seconds. The model uses 3 seconds pending official verification.; The model starts one-second pulses at combat time 1, refreshes the effect on impact instead of multiplying it, samples AP and maximum Health on each pulse and uses provisional adjacency geometry. The video confirms adjacent effects and the multiplier change only.

Alistar: Alistar's official spell data is unavailable. The three star-scaled value sets and 8% maximum Health component use provisional values cross-checked between Lolchess and TFTips.; The video caption says knockup, while current Lolchess text says a 1.5-second stun. The model follows the current text and does not label it a knockup; official verification is still required.; The third-party phrase all impairing effects has no official enumeration. The model clears external crowd control only, without also clearing Burn, Wound, Attack Speed reduction or Stasis. Retargeting if the current target dies before impact remains unverified.

Krug: Krug's official spell values are unavailable. Maximum Health, damage and split-unit Health by star level use the same-day tft.tools 18.2 snapshot and are not classified as official.; The tft.tools damage breakdown uses live maximum Health at cast time before growth. Whether the server increases Health first and then samples the 8% component in the same tick still needs direct script or frame-by-frame Health evidence.; Baby Krug AD, Attack Speed, resistances and attack-role values have no official source. Attack damage is not guessed; the model implements living Health, spawn cells, targetability and taunt.; Slate Buff's 8% maximum Health shield for allies when Krug or a Baby Krug dies is conditional. It is not applied to the base ability and awaits explicit Slate Mark state.; The video verifies the rolling collision only. Death splitting is supported by Riot's overview; taunt duration and official tie-breaking between equally distant targets still need direct data.

Sett: Sett's official Set 18 spell data is unavailable. All current ability values remain provisional, cross-checked between Lolchess and TFTips.; The 90-degree angle, 3-hex range and maximum of five targets are inferred from description keywords, not precise measurements from video or official data. They remain approximate.; The low-Health threshold boundary, charge timing, segmented healing animation and Mana-lock exceptions need official data. The original audit does not classify the ability as exact or enable the Set 18 combat flag before that data is complete.

Gnar: Official leap selection, arc, collision half-width and exact landing timing lack binary data. The model uses legal landing cells within 3 hexes and approximates line collision from the thrown unit's origin to the farthest enemy.; Star-scaled coefficients use cross-checked provisional values and must not be labeled official exact values.

Veigar: Veigar's official spell data is unavailable. Both star-scaled damage sets, the 30% threshold and 1.5% AP per ability kill use provisional values cross-checked between Lolchess and TFTips.; The video directly confirms one primary-target hit. Captions support the branch behavior but not exact coefficients, threshold sampling or projectile speed. The model samples Health percentage on impact and uses a default 0.3-second impact delay.; Repeated casts within one fight are implemented. Permanent carryover between rounds requires full-game player state and is not connected to the single-combat engine.

Ahri: Official ability data is unavailable. Lolchess and TFTips agree on 485/735/3500; the full set of 1.75 seconds, 3 hexes, 20% and the 20% minimum is mainly from TFTips and remains provisional.; Third-party text does not reveal tie-breaking between clusters with equal hit counts. The model chooses the closer cluster, then uses stable row/column ordering pending official verification.; The video does not establish whether the blast center is selected at cast start or impact. The model locks it on cast start rather than automatically tracking a new cluster at impact.

LeBlanc: LeBlanc's official spell data is unavailable. Lolchess and TFTips agree on primary damage [250,375,565], adjacent damage [85,130,190] and base copy chance [10,15,40]; only Lolchess provides complete text for the additional 4% per takedown.; The 71–74-second video segment shows a copy after combat, not the active clone hit. It does not verify targeting, a 1-hex radius, simultaneous impacts or exact coefficients.; Post-combat copying requires strongest-LeBlanc selection, accumulated takedowns, a success roll, eligible targets, bench/board capacity and persistent state. These full-game rules are not connected to single combat. Retargeting before impact and exact timing between the two hits also need official verification.

8 units · 1 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Summoner: Yorick summon Health, Azir soldier damage and Raptor damage are connected. Zyra's +4/+6 attacks are included in the 12-second plant lifetime. Official attack intervals and spawn-selection rules are unavailable, so timing and positions remain provisional video-calibrated assumptions.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Ezreal: Ezreal's official spell values are unavailable. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Blink away has no exact published distance. The model allows up to 2 hexes and selects an available cell that maximizes distance from the nearest enemy, pending official or measurable video evidence.; The current English text does not say instead. The fourth cast therefore keeps its primary single-target damage and adds the wave. If official scripts show replacement behavior, the normal hit must be skipped.; The 8-hex range, 0.65-hex half-width and preference for the current target when hit counts tie are provisional geometry and tie-breaking, not official specifications.

Soraka: The supplied video does not show Soraka; footage of other champions is not used to infer her behavior.; Soraka's official Set 18 spell data is unavailable. Values, impact timing and exact visual spacing of the three extra stars still need official verification.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Azir: Azir's official Set 18 spell data is unavailable. The 150%, six attacks, two soldiers and current Lolchess values [40,60,96] remain provisional and conflict with older tft.tools values.; The older video caption says 6 seconds, while current Lolchess text says the next six attacks. The model follows the newer attack-count limit and retains the version difference in the audit.; Independent soldier crits, exact delay between damage packets, retargeting after the first packet kills and attack resets need official verification. Both packets currently lock the original target and share one normal-attack crit result.; The video does not establish soldiers as independent summons with Health, targetability or autonomous attacks. The model implements the explicitly described command damage only.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Yunara: The split now hits the nearest two enemies within 2 hexes of the primary target, replacing a 1-hex radius that often found only one. This follows the September 4, 2026 report; official text says nearby without a distance, so the radius remains provisional.; Yunara's official spell data is unavailable. Total primary damage [160,240,360] and the 35% split remain provisional. Based on Lolchess's AD/AP icons and variable magnitudes, the model maps ADDamage2 [150,225,335] to AD and ADDamage [10,15,25] to AP; official verification is still needed.; The video shows a short dash without a coordinate grid. The model moves one hex toward the locked target into an unoccupied cell. Official dash distance, passing through units and behavior with no valid forward cell are unknown.; The exact delay from the primary hit to the split is not visible. The model uses 0.1 seconds to preserve the hit-then-split order. The values divide evenly into [56,84,126], but no intermediate display rounding is applied pending confirmation of server rounding.

Yorick: CommunityDragon has character data for Yorick and his Spirit Walker, but the aggregate Set 18 snapshot contained only two shop champions. Active damage, healing and summon Health formulas still use provisional Lolchess values pending verification.; Spirit Walker data does not provide usable base AD, Armor or role classification. Unknown AD and Armor remain null in the source and use a neutral zero in combat; they are not inferred from the video's 27 damage.; The video confirms the order of the single-target active and the summon on death. Full summon attack rules, lifetime, inherited effects and retargeting when the target dies before impact remain unverified.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

8 units · 2 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Summoner: Yorick summon Health, Azir soldier damage and Raptor damage are connected. Zyra's +4/+6 attacks are included in the 12-second plant lifetime. Official attack intervals and spawn-selection rules are unavailable, so timing and positions remain provisional video-calibrated assumptions.

Yunara: The split now hits the nearest two enemies within 2 hexes of the primary target, replacing a 1-hex radius that often found only one. This follows the September 4, 2026 report; official text says nearby without a distance, so the radius remains provisional.; Yunara's official spell data is unavailable. Total primary damage [160,240,360] and the 35% split remain provisional. Based on Lolchess's AD/AP icons and variable magnitudes, the model maps ADDamage2 [150,225,335] to AD and ADDamage [10,15,25] to AP; official verification is still needed.; The video shows a short dash without a coordinate grid. The model moves one hex toward the locked target into an unoccupied cell. Official dash distance, passing through units and behavior with no valid forward cell are unknown.; The exact delay from the primary hit to the split is not visible. The model uses 0.1 seconds to preserve the hit-then-split order. The values divide evenly into [56,84,126], but no intermediate display rounding is applied pending confirmation of server rounding.

Azir: Azir's official Set 18 spell data is unavailable. The 150%, six attacks, two soldiers and current Lolchess values [40,60,96] remain provisional and conflict with older tft.tools values.; The older video caption says 6 seconds, while current Lolchess text says the next six attacks. The model follows the newer attack-count limit and retains the version difference in the audit.; Independent soldier crits, exact delay between damage packets, retargeting after the first packet kills and attack resets need official verification. Both packets currently lock the original target and share one normal-attack crit result.; The video does not establish soldiers as independent summons with Health, targetability or autonomous attacks. The model implements the explicitly described command damage only.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Ezreal: Ezreal's official spell values are unavailable. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Blink away has no exact published distance. The model allows up to 2 hexes and selects an available cell that maximizes distance from the nearest enemy, pending official or measurable video evidence.; The current English text does not say instead. The fourth cast therefore keeps its primary single-target damage and adds the wave. If official scripts show replacement behavior, the normal hit must be skipped.; The 8-hex range, 0.65-hex half-width and preference for the current target when hit counts tie are provisional geometry and tie-breaking, not official specifications.

Soraka: The supplied video does not show Soraka; footage of other champions is not used to infer her behavior.; Soraka's official Set 18 spell data is unavailable. Values, impact timing and exact visual spacing of the three extra stars still need official verification.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Yorick: CommunityDragon has character data for Yorick and his Spirit Walker, but the aggregate Set 18 snapshot contained only two shop champions. Active damage, healing and summon Health formulas still use provisional Lolchess values pending verification.; Spirit Walker data does not provide usable base AD, Armor or role classification. Unknown AD and Armor remain null in the source and use a neutral zero in combat; they are not inferred from the video's 27 damage.; The video confirms the order of the single-target active and the summon on death. Full summon attack rules, lifetime, inherited effects and retargeting when the target dies before impact remain unverified.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

6 units · 2 stars · 0 augments

Primal: — · Pixies 0

Solar: opening shields, shield and bonus magic-damage growth per unique 3-star ally, team Attack Speed/resistances at three 3-stars and conversion of 40% of bonus magic damage to true damage at five are implemented. Randomly upgrading a 3-star to 4-star every 4 seconds after reaching eight 3-stars is missing because 4-star ability data remains incomplete.

Kayle: Direction, area and maximum target count of the 3-star attack wave; Official ordering of Magic Resist reduction and bonus magic damage within one attack

Sejuani: Sejuani's official spell data is unavailable. Shields, the 10% maximum Health component, cone damage and line damage by star level use provisional values cross-checked between Lolchess and TFTips.; Video and text confirm a cone and a line but not precisely a 90-degree angle, 3-hex range, 0.5-hex half-width or 5-target cap. These remain provisional geometry pending official verification.; The delay between stages, reorientation after the target dies, shield-refresh merging and whether the line starts at the cone's end are unknown. The model uses 0.1 seconds, locks direction on cast, adds shields normally and uses the caster as the origin for both stages.

Leona: Leona's official Set 18 spell data is unavailable. Starting resistances, Armor scaling and stun values remain provisional, cross-checked between Lolchess and TFTips.; The video confirms the visible decay, target and stun but cannot establish the precise AP snapshot time. The model uses the usual combat-start snapshot.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Alune: Official data has not populated NumMoonshards, NumEnemies or MagicDamageCalc1/2.; Normal and full-moon coefficients by star level remain provisional.; The available direct binary payload conflicts with the current-set ability and cannot be promoted to verified data.

9 units · 1 stars · 0 augments

Primal: — · Pixies 0

Durability may be underestimated for units that reach the 15-second survival cap.

Fae: with an explicit Pixie count, each Pixie grants Fae units 5%/8% AD and AP, plus a one-time heal below 50% Health for 2.5%/4% maximum Health per Pixie. Little Buddy provides attraction thresholds and Golden Pixie rewards. Automatic damage/heal/shield accumulation and cross-combat gold settlement are not connected.

Elderwood: three separate combat entities, the total-star panel, death stun, periodic AD/AP, Deepwood damage and healing/Attack Speed on ally death are connected. Values use the September 5, 2026 TFTips snapshot. Additional rules at 11 units, secondary-hit geometry and healing scaling need official calibration.

Gnar: Official leap selection, arc, collision half-width and exact landing timing lack binary data. The model uses legal landing cells within 3 hexes and approximates line collision from the thrown unit's origin to the farthest enemy.; Star-scaled coefficients use cross-checked provisional values and must not be labeled official exact values.

Ezreal: Ezreal's official spell values are unavailable. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Blink away has no exact published distance. The model allows up to 2 hexes and selects an available cell that maximizes distance from the nearest enemy, pending official or measurable video evidence.; The current English text does not say instead. The fourth cast therefore keeps its primary single-target damage and adds the wave. If official scripts show replacement behavior, the normal hit must be skipped.; The 8-hex range, 0.65-hex half-width and preference for the current target when hit counts tie are provisional geometry and tie-breaking, not official specifications.

Hecarim: Official ability data is unavailable. Healing [375,475,685], magic damage [80,120,195], stun duration [1.5,1.5,1.75], 50 resistances, 3 seconds and the nearest three enemies agree across two sources but remain provisional.; The video confirms visible multi-target hits and healing, not exact values, a three-target limit, control type, resistance values or a 3-second duration.; Third-party sources do not specify the spectral rider's flight time, collision rules or cast windup. The model uses a default 0.3-second cast, locks targets on cast start and resolves impacts together; this is not verified server timing.

Alistar: Alistar's official spell data is unavailable. The three star-scaled value sets and 8% maximum Health component use provisional values cross-checked between Lolchess and TFTips.; The video caption says knockup, while current Lolchess text says a 1.5-second stun. The model follows the current text and does not label it a knockup; official verification is still required.; The third-party phrase all impairing effects has no official enumeration. The model clears external crowd control only, without also clearing Burn, Wound, Attack Speed reduction or Stasis. Retargeting if the current target dies before impact remains unverified.

LeBlanc: LeBlanc's official spell data is unavailable. Lolchess and TFTips agree on primary damage [250,375,565], adjacent damage [85,130,190] and base copy chance [10,15,40]; only Lolchess provides complete text for the additional 4% per takedown.; The 71–74-second video segment shows a copy after combat, not the active clone hit. It does not verify targeting, a 1-hex radius, simultaneous impacts or exact coefficients.; Post-combat copying requires strongest-LeBlanc selection, accumulated takedowns, a success roll, eligible targets, bench/board capacity and persistent state. These full-game rules are not connected to single combat. Retargeting before impact and exact timing between the two hits also need official verification.

Ornn: The cone now has a 1-hex range, 120-degree angle and maximum of 3 targets, following the September 4, 2026 screenshot report that the breath mainly hits the unit directly ahead. Official text gives no range, so the geometry remains provisional.; Official ability variables are unavailable. Two sources agree on 400/460/550 shields and 200/300/450 damage, but the values remain provisional.; The video confirms a short-range cone but cannot establish a 90-degree angle, 3-hex range or 5-target limit precisely. These remain provisional geometry.; The Artifact Anvil reward at 90,000 accumulated damage taken, with doubled progress at 3 stars, is a cross-combat economy quest and is not connected to the combat engine.

Xayah: Xayah's official Set 18 spell data is unavailable. The values 72/108/165, 50%, 5 attacks and 2 Armor reduction per attack remain provisional, cross-checked against third-party text and video captions.; The video shows feather paths but does not establish piercing or damage to secondary targets. The model follows the text and hits the current attack target without adding unsupported area damage.; Official data does not yet specify attack resets, missile speed or exact travel time.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Lillia: Lillia's official ability data is unavailable. Healing [325,475,800], magic damage [90,135,2000], sleep [1.5,1.75,8], four butterflies, the 1,000-damage threshold and 10% maximum Health bonus damage agree across two sources but remain provisional.; The captions at 266–269 seconds show an older five-target / 2-second version that conflicts with current sources. The video supports visible behavior, not the current target count or duration.; Sources do not specify whether the initial hit counts toward 1,000, whether the threshold is before/after mitigation or shields, whether natural waking adds damage, or whether butterflies can repeat targets. The model counts the initial hit after mitigation but before shields, adds no damage on natural expiry and chooses the nearest four unique enemies.

Rammus: Official taunt radius and duration; the current assumptions are 2 hexes and 4 seconds; Official calculation of shield-break damage with ADDamageRatio 0.5

9 units · 1 stars · 0 augments

Primal: — · Pixies 0

Despoiler: damage resolution is fully connected in combat. The older one-sided base-DPS preview lacks an enemy Health pool and can apply only the base amplification.

Riftbeast: shop occupation every three combats at five units, the planned +2 team-size effect at ten and combat growth at seven are implemented. At three, the selected Alpha Mark applies color buffs for nine Riftbeasts; Burn and shield durations remain provisional. Full-game shop and team-size application are still pending.

Emerald Avatar: explicit partner selection by unit ID or cell, healing for both units, two empowered attacks each and a 3-hex shield from each position the first time either drops below 50% Health are implemented. Overlapping allies receive one shield. Official coefficients and same-tick ordering for lethal threshold-crossing damage remain unverified.

Alune: Official data has not populated NumMoonshards, NumEnemies or MagicDamageCalc1/2.; Normal and full-moon coefficients by star level remain provisional.; The available direct binary payload conflicts with the current-set ability and cannot be promoted to verified data.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Diana: Diana's official spell data is unavailable. Six orbs, 2 hexes, damage per orb [65,100,155], shields [150,225,300] and 2 seconds agree across two sources but remain provisional.; Sources do not specify whether six orbs are distributed evenly, randomly or by other weights among valid enemies. The model assigns each orb independently and records the decisions; this is not an official selection rule.; A clear current-version video, precise projectile travel, behavior after death before impact and independent expiry of multiple shields are unverified. The model uses a default 0.3-second impact delay and shared shield expiry.

Taric: Official Set 18 ability data is unavailable for field-by-field verification. Exact values remain provisional, cross-checked between the same-day third-party sources.; For a single lethal hit that also crosses 50% Health, the official order of shield application and damage needs frame-by-frame or binary-event evidence. The model triggers at the next 0.1-second semantic check after impact.; The 0.4-second cast is a provisional default. The video confirms effect order, not exact windup or recovery.

Brambleback: Red Buff variant: attacks apply 1% Burn and heal 4% maximum Health; Official leap landing rule; the model uses the nearest empty cell beside the target

Sentinel: Ancient Sentinel's official spell values are unavailable. Shield, maximum Health scaling, damage and 10 Mana Reave use provisional values from the current tft.tools/Lolchess snapshot.; The ability says a brief knockup without a duration. The engine does not invent 0.1, 0.25 or another duration that would change combat scheduling.; The fissure uses an inferred 8-hex length and 0.5-hex half-width based on the board-wide description. Text supports aiming at the largest group, but exact collision width requires official data or direct video evidence.; Blue Buff's 2 Mana Regen for allies per cast applies only when Sentinel receives the matching Alpha/Blue bonus. That selection state is not connected to the unit, so it is excluded from the base ability.

Mama Beak: Official character data identifies the mother and baby Raptors, but the ability is still a generic placeholder. Four birds, 5 seconds and damage [27,41,65] agree across two third-party pages but remain provisional.; The video confirms separate visible models and coordinated attacks, but not spawn cells, movement, travel time or collision. The model uses synchronized damage proxies that do not occupy board cells.; Summon crits, on-hit/item triggers, hits after the mother's attack kills the target, and stacking versus refreshing on recast lack official evidence. The model currently uses no crits, no triggers, no retargeting and refreshes.; The 0.25-second cast uses the shared spell placeholder's timing. It is not a calibrated official windup.; The source and stacking rules of the orange buff are unpublished. Armor reduction is applied only when that external effect is selected.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Murkwolf: Murk Wolf's official spell values are unavailable. The three damage sets, 175% Attack Speed and two-attack duration come from tft.tools/Lolchess-derived data and are not classified as official.; Healing after impact in the video is absent from the current 18.2 description and is treated as old behavior. Current damage is not inferred from older floating numbers.; Grey Buff's Precision and missing-Health crit bonus require a conditional Alpha Mark. Its acquisition state is not connected to the unit, so the base ability does not receive it automatically.; The model interprets lowest Health as absolute current Health, following the English text. If official scripts use Health percentage, the selector and fixed trajectory must be updated.

Pebbles: CommunityDragon has not supplied usable official spell data for Pebbles. Damage by star level, Mana drain and Magic Resist reduction still use provisional values.; Teal Buff requires an explicit Alpha Riftbeast selection. The board controls do not expose that selection yet, so the bonus of 2 Mana Regen per 4 seconds of channeling is not applied.; Whether crowd control cancels or pauses the channel, and how untargetability affects it, still need official scripts or frame-by-frame evidence. The current model ends the channel when crowd control is applied or the locked target dies.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Despoiler: damage resolution is fully connected in combat. The older one-sided base-DPS preview lacks an enemy Health pool and can apply only the base amplification.

Riftbeast: shop occupation every three combats at five units, the planned +2 team-size effect at ten and combat growth at seven are implemented. At three, the selected Alpha Mark applies color buffs for nine Riftbeasts; Burn and shield durations remain provisional. Full-game shop and team-size application are still pending.

Emerald Avatar: explicit partner selection by unit ID or cell, healing for both units, two empowered attacks each and a 3-hex shield from each position the first time either drops below 50% Health are implemented. Overlapping allies receive one shield. Official coefficients and same-tick ordering for lethal threshold-crossing damage remain unverified.

Diana: Diana's official spell data is unavailable. Six orbs, 2 hexes, damage per orb [65,100,155], shields [150,225,300] and 2 seconds agree across two sources but remain provisional.; Sources do not specify whether six orbs are distributed evenly, randomly or by other weights among valid enemies. The model assigns each orb independently and records the decisions; this is not an official selection rule.; A clear current-version video, precise projectile travel, behavior after death before impact and independent expiry of multiple shields are unverified. The model uses a default 0.3-second impact delay and shared shield expiry.

Aphelios: Aphelios's official ability script is unavailable. Star-scaled coefficients, 2 seconds, one extra attack per 20% bonus Attack Speed and a 2-hex radius agree between current tft.tools and Lolchess but remain provisional.; Public text says the burst is split among enemies within 2 hexes but does not establish whether the center is the target's cast-time position, end position or projectile impact. The model fixes the center on cast.; The video lacks a complete combat board and cannot verify the AD/AP split of the 450 and 40 components. The model follows Lolchess's variable decomposition of 450% AD plus 40% AP pending official verification.; Whether crowd control cancels the final burst, whether the last strike and burst share a tick, and whether remaining strikes transfer after the target dies are unknown. The model allows interruption, keeps target lock and processes events in order at the same semantic boundary.

Alune: Official data has not populated NumMoonshards, NumEnemies or MagicDamageCalc1/2.; Normal and full-moon coefficients by star level remain provisional.; The available direct binary payload conflicts with the current-set ability and cannot be promoted to verified data.

Taric: Official Set 18 ability data is unavailable for field-by-field verification. Exact values remain provisional, cross-checked between the same-day third-party sources.; For a single lethal hit that also crosses 50% Health, the official order of shield application and damage needs frame-by-frame or binary-event evidence. The model triggers at the next 0.1-second semantic check after impact.; The 0.4-second cast is a provisional default. The video confirms effect order, not exact windup or recovery.

Brambleback: Red Buff variant: attacks apply 1% Burn and heal 4% maximum Health; Official leap landing rule; the model uses the nearest empty cell beside the target

Sentinel: Ancient Sentinel's official spell values are unavailable. Shield, maximum Health scaling, damage and 10 Mana Reave use provisional values from the current tft.tools/Lolchess snapshot.; The ability says a brief knockup without a duration. The engine does not invent 0.1, 0.25 or another duration that would change combat scheduling.; The fissure uses an inferred 8-hex length and 0.5-hex half-width based on the board-wide description. Text supports aiming at the largest group, but exact collision width requires official data or direct video evidence.; Blue Buff's 2 Mana Regen for allies per cast applies only when Sentinel receives the matching Alpha/Blue bonus. That selection state is not connected to the unit, so it is excluded from the base ability.

Mama Beak: Official character data identifies the mother and baby Raptors, but the ability is still a generic placeholder. Four birds, 5 seconds and damage [27,41,65] agree across two third-party pages but remain provisional.; The video confirms separate visible models and coordinated attacks, but not spawn cells, movement, travel time or collision. The model uses synchronized damage proxies that do not occupy board cells.; Summon crits, on-hit/item triggers, hits after the mother's attack kills the target, and stacking versus refreshing on recast lack official evidence. The model currently uses no crits, no triggers, no retargeting and refreshes.; The 0.25-second cast uses the shared spell placeholder's timing. It is not a calibrated official windup.; The source and stacking rules of the orange buff are unpublished. Armor reduction is applied only when that external effect is selected.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Murkwolf: Murk Wolf's official spell values are unavailable. The three damage sets, 175% Attack Speed and two-attack duration come from tft.tools/Lolchess-derived data and are not classified as official.; Healing after impact in the video is absent from the current 18.2 description and is treated as old behavior. Current damage is not inferred from older floating numbers.; Grey Buff's Precision and missing-Health crit bonus require a conditional Alpha Mark. Its acquisition state is not connected to the unit, so the base ability does not receive it automatically.; The model interprets lowest Health as absolute current Health, following the English text. If official scripts use Health percentage, the selector and fixed trajectory must be updated.

Pebbles: CommunityDragon has not supplied usable official spell data for Pebbles. Damage by star level, Mana drain and Magic Resist reduction still use provisional values.; Teal Buff requires an explicit Alpha Riftbeast selection. The board controls do not expose that selection yet, so the bonus of 2 Mana Regen per 4 seconds of channeling is not applied.; Whether crowd control cancels or pauses the channel, and how untargetability affects it, still need official scripts or frame-by-frame evidence. The current model ends the channel when crowd control is applied or the locked target dies.

7 units · 1 stars · 0 augments

Primal: — · Pixies 0

Infernal: combat Burn/Wound and the post-combat plan to replace 1/2/4 non-Infernal shop slots with units costing one more gold at 3/5/7 are implemented. Applying these changes to full shop/pool state is still pending.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Amumu: Amumu's official spell values are unavailable. Star-scaled coefficients, the 2.2% maximum Health component, radius and bonus stun against burning enemies come from the same-day tft.tools 18.2 snapshot and remain provisional.; Lolchess's Korean description still lists a total stun of 3/3/13 seconds against burning enemies, conflicting with tft.tools's extra 0.5/0.5/3 seconds. The model follows the newer source explicitly dated for 18.2 and retains the older values as a documented conflict.; The video's 1.25-second caption and similar values are from an older version and support behavior only. The first passive pulse, same-tick ordering of active damage and control, and whether newly applied Burn can extend that same stun need official verification.; The burning condition currently accepts only item or trait Burn already present before impact. Amumu's own active is not assumed to apply Burn first, and Infernal membership does not guarantee that the target is burning.

Vi: Vi's official Set 18 spell data is unavailable. The current values are third-party provisional data; the older TFTips snapshot is retained only as evidence of a version difference.; The video text says 6 seconds while current Lolchess says 3 seconds. The model follows the newer 3 seconds until official data resolves the version difference.; The video cannot verify all crowd-control immunity exceptions involving displacement, knockups or special forced controls. The model uses normal crowd-control immunity.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Sejuani: Sejuani's official spell data is unavailable. Shields, the 10% maximum Health component, cone damage and line damage by star level use provisional values cross-checked between Lolchess and TFTips.; Video and text confirm a cone and a line but not precisely a 90-degree angle, 3-hex range, 0.5-hex half-width or 5-target cap. These remain provisional geometry pending official verification.; The delay between stages, reorientation after the target dies, shield-refresh merging and whether the line starts at the cone's end are unknown. The model uses 0.1 seconds, locks direction on cast, adds shields normally and uses the caster as the origin for both stages.

Scuttlecrab: Scuttle's official Set 18 spell data is unavailable. Damage, healing, 15% and 3 seconds use provisional values cross-checked between Lolchess and TFTips.; The video does not establish whether burrowing grants untargetability, moves the unit, interrupts attacks or changes the first healing tick. The model keeps the unit targetable and starts one-second healing intervals immediately on impact.; Green Buff's source, repeat triggers, per-unit cooldown and ally rules require Riftbeast selection state or official data. The effect is currently not applied.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

Kennen: Kennen's official ability data is still a generic placeholder. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The video confirms charging, dashing into a cluster and persistent fire at the landing point, but not exact cluster selection, dash width, landing avoidance or travel time. A 2-hex cluster, 0.65-hex half-width, nearest empty cluster cell, mirrored tie-break and 0.3-second hop are provisional assumptions.; Text provides total damage over 2 seconds without a server tick rate. The model uses four ticks at 0.5-second intervals while preserving the total damage pool.; Whether Burn includes every special mark, dead/untargetable enemies and a snapshot at charge start or end needs official scripts. The model counts living enemies with active item or trait Burn at cast start.

Ezreal: Ezreal's official spell values are unavailable. Star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; Blink away has no exact published distance. The model allows up to 2 hexes and selects an available cell that maximizes distance from the nearest enemy, pending official or measurable video evidence.; The current English text does not say instead. The fourth cast therefore keeps its primary single-target damage and adds the wave. If official scripts show replacement behavior, the normal hit must be skipped.; The 8-hex range, 0.65-hex half-width and preference for the current target when hit counts tie are provisional geometry and tie-breaking, not official specifications.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Summoner: Yorick summon Health, Azir soldier damage and Raptor damage are connected. Zyra's +4/+6 attacks are included in the 12-second plant lifetime. Official attack intervals and spawn-selection rules are unavailable, so timing and positions remain provisional video-calibrated assumptions.

Yorick: CommunityDragon has character data for Yorick and his Spirit Walker, but the aggregate Set 18 snapshot contained only two shop champions. Active damage, healing and summon Health formulas still use provisional Lolchess values pending verification.; Spirit Walker data does not provide usable base AD, Armor or role classification. Unknown AD and Armor remain null in the source and use a neutral zero in combat; they are not inferred from the video's 27 damage.; The video confirms the order of the single-target active and the summon on death. Full summon attack rules, lifetime, inherited effects and retargeting when the target dies before impact remain unverified.

Karma: Official ability data is unavailable. Two sources agreed on 280/420/630 total tether damage and 120/180/270 burst damage, but these remain third-party provisional values.; The video confirms the tether and secondary burst, but not server damage timing. The current 0.5-second damage interval is a calculation assumption that needs official verification.; Behavior when the tether target dies, becomes untargetable or moves before the tether ends is unverified: cancellation, an explosion at the original location and tracking are all unresolved. Only a trajectory with a surviving target has been verified.

Yunara: The split now hits the nearest two enemies within 2 hexes of the primary target, replacing a 1-hex radius that often found only one. This follows the September 4, 2026 report; official text says nearby without a distance, so the radius remains provisional.; Yunara's official spell data is unavailable. Total primary damage [160,240,360] and the 35% split remain provisional. Based on Lolchess's AD/AP icons and variable magnitudes, the model maps ADDamage2 [150,225,335] to AD and ADDamage [10,15,25] to AP; official verification is still needed.; The video shows a short dash without a coordinate grid. The model moves one hex toward the locked target into an unoccupied cell. Official dash distance, passing through units and behavior with no valid forward cell are unknown.; The exact delay from the primary hit to the split is not visible. The model uses 0.1 seconds to preserve the hit-then-split order. The values divide evenly into [56,84,126], but no intermediate display rounding is applied pending confirmation of server rounding.

Master Yi: Official priority when AD and AP are equal

Ahri: Official ability data is unavailable. Lolchess and TFTips agree on 485/735/3500; the full set of 1.75 seconds, 3 hexes, 20% and the 20% minimum is mainly from TFTips and remains provisional.; Third-party text does not reveal tie-breaking between clusters with equal hit counts. The model chooses the closer cluster, then uses stable row/column ordering pending official verification.; The video does not establish whether the blast center is selected at cast start or impact. The model locks it on cast start rather than automatically tracking a new cluster at impact.

Sett: Sett's official Set 18 spell data is unavailable. All current ability values remain provisional, cross-checked between Lolchess and TFTips.; The 90-degree angle, 3-hex range and maximum of five targets are inferred from description keywords, not precise measurements from video or official data. They remain approximate.; The low-Health threshold boundary, charge timing, segmented healing animation and Mana-lock exceptions need official data. The original audit does not classify the ability as exact or enable the Set 18 combat flag before that data is complete.

Ashe: Ashe's official spell values are unavailable. All star-scaled coefficients come from the current tft.tools/Lolchess snapshot and remain provisional.; The 8-hex length and 0.65-hex half-width are inferred from a board-piercing description, not official geometry. Exact collision width still needs official data.; The 20% Slow applies while enemies are inside the path. The engine approximates a continuous area with one-second ticks and a 0.05-second overlap refresh; actual server frequency is unknown.; The video confirms a piercing line and frost path but cannot independently establish tie-breaking when aiming at the most enemies. Direction follows the current ability text.

Maokai: Maokai's official ability data is still a generic placeholder. Star-scaled values come from the current tft.tools/Lolchess snapshot and remain provisional.; Riot has not defined the server counter for damage blocked. The model counts raw damage minus damage after resistances, Durability and flat reductions, without counting shields twice. This must be checked against official scripts.; Sapling target selection, jump speed, simultaneous assignment and stat snapshots after the source dies are unpublished. The model launches them sequentially at the nearest enemy, 0.1 seconds apart.; The video confirms the active effect and healing but not precise sapling spawn positions, hit radius or interception. Playback draws a jump from the source cell to the target cell.

Azir: Azir's official Set 18 spell data is unavailable. The 150%, six attacks, two soldiers and current Lolchess values [40,60,96] remain provisional and conflict with older tft.tools values.; The older video caption says 6 seconds, while current Lolchess text says the next six attacks. The model follows the newer attack-count limit and retains the version difference in the audit.; Independent soldier crits, exact delay between damage packets, retargeting after the first packet kills and attack resets need official verification. Both packets currently lock the original target and share one normal-attack crit result.; The video does not establish soldiers as independent summons with Health, targetability or autonomous attacks. The model implements the explicitly described command damage only.; The original audit does not classify this as exact or enable the Set 18 combat flag until complete official data is available.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Transfigurist: the 25%/35%/50% bonus, AD/AP forms for five champions, Akali's bonus damage against burning enemies and recast on kill, Gromp's Purple Alpha Buff and Nidalee's third tracking spear are implemented. Equal same-scale bonuses use the roster's default form; official comparison/tie rules and unpublished geometry or coefficients remain unresolved.

Riftbeast: shop occupation every three combats at five units, the planned +2 team-size effect at ten and combat growth at seven are implemented. At three, the selected Alpha Mark applies color buffs for nine Riftbeasts; Burn and shield durations remain provisional. Full-game shop and team-size application are still pending.

Pebbles: CommunityDragon has not supplied usable official spell data for Pebbles. Damage by star level, Mana drain and Magic Resist reduction still use provisional values.; Teal Buff requires an explicit Alpha Riftbeast selection. The board controls do not expose that selection yet, so the bonus of 2 Mana Regen per 4 seconds of channeling is not applied.; Whether crowd control cancels or pauses the channel, and how untargetability affects it, still need official scripts or frame-by-frame evidence. The current model ends the channel when crowd control is applied or the locked target dies.

Teemo: Official Set 18 spell values are unavailable for field-by-field verification. The values 60/90/135 remain provisional.; The short video confirms mushrooms hitting multiple targets, but not projectile speed, hit radius or extra splash damage.; The 10%/12%/15% mushroom gathering chance and reroll, player Health and XP rewards are cross-combat economy events. They appear in descriptions but are not rolled during combat.

Kog'Maw: official values for the AP form's damage over time

Morgana: Morgana's official ability values are unavailable. Only the same-day Lolchess page fully lists the three value sets, 4 seconds, nearest three enemies and 25% Omnivamp, so these remain provisional.; Whether the area stays fixed or follows the caster is unverified. It currently stays at the cast location; following behavior is not claimed as verified.; Ticks occur on complete second boundaries from impact +1 through impact +4. The 0.3-second windup, first-tick timing and ordering with expiry still need official scripts.

Sentinel: Ancient Sentinel's official spell values are unavailable. Shield, maximum Health scaling, damage and 10 Mana Reave use provisional values from the current tft.tools/Lolchess snapshot.; The ability says a brief knockup without a duration. The engine does not invent 0.1, 0.25 or another duration that would change combat scheduling.; The fissure uses an inferred 8-hex length and 0.5-hex half-width based on the board-wide description. Text supports aiming at the largest group, but exact collision width requires official data or direct video evidence.; Blue Buff's 2 Mana Regen for allies per cast applies only when Sentinel receives the matching Alpha/Blue bonus. That selection state is not connected to the unit, so it is excluded from the base ability.

Elise: Complete official spell data is still unavailable; the values remain provisional.; The transformation animation does not itself change numerical results. Impact timing must be recalibrated if official data establishes a different animation duration.

Gromp: Official ability data is unavailable. Lolchess and TFTips agree on 250/375/565 direct damage and 100/240/360 total area damage, but the values remain provisional.; The video cannot establish exact projectile speed, collision radius or damage-tick timing. The model assumes 8 hexes per second, a 0.9 world-space collision radius and one-second ticks pending official verification.; The Transfigurist AD branch's damage and combat-entry form selection are connected. Its strong slow is not applied because the exact amount and duration are unknown.; Purple Buff grants 25% Ability Power every 5 seconds through external Riftbeast color state. That state is not connected, so the bonus is not automatically applied to Gromp.; Exact amount and duration of the AD form's strong slow; Complete Purple Buff effect and acquisition conditions; Official priority when AD and AP are equal

Gnar: Official leap selection, arc, collision half-width and exact landing timing lack binary data. The model uses legal landing cells within 3 hexes and approximates line collision from the thrown unit's origin to the farthest enemy.; Star-scaled coefficients use cross-checked provisional values and must not be labeled official exact values.

Hecarim: Official ability data is unavailable. Healing [375,475,685], magic damage [80,120,195], stun duration [1.5,1.5,1.75], 50 resistances, 3 seconds and the nearest three enemies agree across two sources but remain provisional.; The video confirms visible multi-target hits and healing, not exact values, a three-target limit, control type, resistance values or a 3-second duration.; Third-party sources do not specify the spectral rider's flight time, collision rules or cast windup. The model uses a default 0.3-second cast, locks targets on cast start and resolves impacts together; this is not verified server timing.

9 units · 2 stars · 0 augments

Primal: — · Pixies 0

Transfigurist: the 25%/35%/50% bonus, AD/AP forms for five champions, Akali's bonus damage against burning enemies and recast on kill, Gromp's Purple Alpha Buff and Nidalee's third tracking spear are implemented. Equal same-scale bonuses use the roster's default form; official comparison/tie rules and unpublished geometry or coefficients remain unresolved.

Riftbeast: shop occupation every three combats at five units, the planned +2 team-size effect at ten and combat growth at seven are implemented. At three, the selected Alpha Mark applies color buffs for nine Riftbeasts; Burn and shield durations remain provisional. Full-game shop and team-size application are still pending.

Morgana: Morgana's official ability values are unavailable. Only the same-day Lolchess page fully lists the three value sets, 4 seconds, nearest three enemies and 25% Omnivamp, so these remain provisional.; Whether the area stays fixed or follows the caster is unverified. It currently stays at the cast location; following behavior is not claimed as verified.; Ticks occur on complete second boundaries from impact +1 through impact +4. The 0.3-second windup, first-tick timing and ordering with expiry still need official scripts.

Sentinel: Ancient Sentinel's official spell values are unavailable. Shield, maximum Health scaling, damage and 10 Mana Reave use provisional values from the current tft.tools/Lolchess snapshot.; The ability says a brief knockup without a duration. The engine does not invent 0.1, 0.25 or another duration that would change combat scheduling.; The fissure uses an inferred 8-hex length and 0.5-hex half-width based on the board-wide description. Text supports aiming at the largest group, but exact collision width requires official data or direct video evidence.; Blue Buff's 2 Mana Regen for allies per cast applies only when Sentinel receives the matching Alpha/Blue bonus. That selection state is not connected to the unit, so it is excluded from the base ability.

Kog'Maw: official values for the AP form's damage over time

Teemo: Official Set 18 spell values are unavailable for field-by-field verification. The values 60/90/135 remain provisional.; The short video confirms mushrooms hitting multiple targets, but not projectile speed, hit radius or extra splash damage.; The 10%/12%/15% mushroom gathering chance and reroll, player Health and XP rewards are cross-combat economy events. They appear in descriptions but are not rolled during combat.

Pebbles: CommunityDragon has not supplied usable official spell data for Pebbles. Damage by star level, Mana drain and Magic Resist reduction still use provisional values.; Teal Buff requires an explicit Alpha Riftbeast selection. The board controls do not expose that selection yet, so the bonus of 2 Mana Regen per 4 seconds of channeling is not applied.; Whether crowd control cancels or pauses the channel, and how untargetability affects it, still need official scripts or frame-by-frame evidence. The current model ends the channel when crowd control is applied or the locked target dies.

Elise: Complete official spell data is still unavailable; the values remain provisional.; The transformation animation does not itself change numerical results. Impact timing must be recalibrated if official data establishes a different animation duration.

Gromp: Official ability data is unavailable. Lolchess and TFTips agree on 250/375/565 direct damage and 100/240/360 total area damage, but the values remain provisional.; The video cannot establish exact projectile speed, collision radius or damage-tick timing. The model assumes 8 hexes per second, a 0.9 world-space collision radius and one-second ticks pending official verification.; The Transfigurist AD branch's damage and combat-entry form selection are connected. Its strong slow is not applied because the exact amount and duration are unknown.; Purple Buff grants 25% Ability Power every 5 seconds through external Riftbeast color state. That state is not connected, so the bonus is not automatically applied to Gromp.; Exact amount and duration of the AD form's strong slow; Complete Purple Buff effect and acquisition conditions; Official priority when AD and AP are equal

Gnar: Official leap selection, arc, collision half-width and exact landing timing lack binary data. The model uses legal landing cells within 3 hexes and approximates line collision from the thrown unit's origin to the farthest enemy.; Star-scaled coefficients use cross-checked provisional values and must not be labeled official exact values.

Hecarim: Official ability data is unavailable. Healing [375,475,685], magic damage [80,120,195], stun duration [1.5,1.5,1.75], 50 resistances, 3 seconds and the nearest three enemies agree across two sources but remain provisional.; The video confirms visible multi-target hits and healing, not exact values, a three-target limit, control type, resistance values or a 3-second duration.; Third-party sources do not specify the spectral rider's flight time, collision rules or cast windup. The model uses a default 0.3-second cast, locks targets on cast start and resolves impacts together; this is not verified server timing.