Claude Confusions
Known mistakes and misunderstandings Claude has made about Evokers mechanics. Documented here so they are not repeated.
82 confusions logged.
Fusion setup takes 3+ turns
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Said the Seir + Beelzebub fusion required '3+ turns of tempo' to set up. In reality it's 2 turns (1 contract per turn), which is the minimum for any fusion. The setup cost is not a weakness specific to this combo.
Correct rule: Each contract phase plays 1 demon. A fusion requires 2 demons = 2 contract phases = 2 turns. This is the baseline minimum for all fusions, not an extra cost.
Familiars are not demons
| Impact | High |
| Category | Terminology |
| Corrected | Yes |
What went wrong: Assumed familiars and demons are separate types. Concluded that Seir's passive (triggers on demon death) would NOT activate when Locusts (familiars) are fatally wounded via Gluttony. Rated the Seir + Beelzebub fusion at 6/10 instead of 8/10.
Correct rule: Familiars ARE demons. The game uses terms like 'fusable demons' or CP to distinguish between units and familiars — not by categorizing them as different types. Any ability referencing 'demons' includes familiars.
Hand size changes across rest periods
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Assumed hand size and selection narrowed as the game progressed, with fewer speed options available in later rest periods. Built an entire strategic framework around 'saving fast demons across rest periods' and deck depletion affecting choice. In reality, you always draw 1 + hold 2 = choose from exactly 3 cards every cycle, constant throughout the game.
Correct rule: Each cycle: draw 1 card (hand becomes 3), contract 1 (hand returns to 2). You always choose among 3 demons regardless of game state. You cannot strategically 'bank' specific speed demons across rest periods.
Misidentified fusion bonus stats as 'ultimate form' stats
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Labeled UHP, UPWR, and UVP as 'ultimate form' stats and initially claimed they were not gameplay-relevant. These are actually fusion bonus stats printed on the bottom of each card — when a card is used as the bottom card in a fusion, these values are added to the top card's HP, PWR, and VP.
Correct rule: UHP, UPWR, and UVP are fusion bonus modifiers. When two demons fuse, the top card provides base stats and the bottom card's UHP/UPWR/UVP are added as bonuses. These stats are integral to fusion gameplay, not cosmetic or unused.
Over-valued stats, undervalued abilities when ranking demons
| Impact | High |
| Category | Stats |
| Corrected | Yes |
What went wrong: Ranked Zagan, Dantalion, and Purson as the weakest demons based on their shared 6/2/1 stat line. Repeatedly anchored on raw stats (HP, PWR, VP) as the primary measure of power, treating abilities as tiebreakers rather than the defining factor.
Correct rule: In Evokers, abilities define what a demon does. Stats set boundaries, abilities define roles. A 6/2/1 demon with a powerful passive can be far more impactful than a high-stat demon with a weak ability. Always lead analysis with abilities and fusion synergies.
Time Dilation is not a free action
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Described Gaap's Time Dilation as granting a 'free preemptive strike' to an allied demon. Ignored the '(with cost)' clause, which means the allied demon still pays full AP and exhaust for whatever action it performs.
Correct rule: Time Dilation (Start of Turn, 0 AP, 1x) lets a target allied demon perform 1 action WITH COST — the ally pays full AP and exhaust. The only advantage is timing: acting before the opponent during their main phase. It is a one-shot timing trick, not a resource advantage.
Shax chain-kills are practical with PWR:4
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Picked Shax as #3 because self-ready on kill 'breaks the exhaust bottleneck' and enables chain-killing. Failed to check whether PWR:4 can actually secure kills. Most demons have 9-15 HP — Shax needs 3-4 attacks to kill one, but only gets 1 per rest cycle. The snowball requires multiple enemies pre-chipped to ≤4 HP in the same lane, which is unrealistic setup.
Correct rule: Always stress-test setup conditions for conditional abilities. If a trigger requires specific board states (multiple low-HP enemies in one lane), evaluate how realistic that state is before calling the ability strong. Shax's chain-kill is win-more, not a reliable strength.
game_config.json says 'first to 15 CP wins'
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Trusted game_config.json's victory description literally without cross-referencing ability text that implies CP is negative (e.g., Virgo's 'your opponent gains 1 CP' is clearly offensive, not a gift).
Correct rule: The game_config description may be outdated or misleading. CP is a loss condition confirmed by the designer. Always treat CP gain as harmful to the gaining player.
Fear of the Dark is a permanent readiness lock
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Picked Nyx as #3 strongest demon because Fear of the Dark's 'cannot be readied' Status seemed to permanently remove demons from the action economy, accumulating locks each rest cycle. In reality, statuses expire at end of current main phase — Fear of the Dark is a 1-phase stall, same problem as Gaap's Unmoving Time.
Correct rule: All Status effects expire at end of the current main phase. Fear of the Dark's 'cannot be readied' only lasts the phase it was applied. The target readies normally at the next rest phase.
Local lockdown only covers 1 of 3 lanes
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Described Gaap's Unmoving Time as 'the best lockdown in the game' and 'full lane shutdown' without considering that the game has 3 lanes. A Local-range lockdown only affects 1/3 of the board. The opponent can spread demons across lanes to minimize impact.
Correct rule: Local-range abilities only affect demons in the same lane. With 3 lanes, opponents can position around local lockdowns. Unmoving Time (2 AP + exhaust) locks one lane per rest cycle. Evaluate local abilities against the 3-lane board, not in isolation.
CP is a win condition (reach 15 to win)
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Treated CP as a victory score — ranked Baal and Beleth as the two strongest demons because their abilities generate CP and they have VP:6. In reality CP is corruption: you LOSE at 15. This made Baal/Beleth's self-corrupting abilities and high VP massive liabilities, not strengths.
Correct rule: CP (Corruption Points) is a loss condition. You lose when you reach 15 CP. When your demon dies, you gain its VP as CP. Abilities that say 'Gain X CP' self-corrupt the user. High VP demons are risky to field because dying costs you more CP.
AP vs Exhaust bottleneck
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Overvalued AP cost reduction (e.g. Murmur's -2 AP) as game-warping, treating it as enabling unlimited attacks. Failed to recognize that exhaust/ready is the primary constraint on combat output, not AP. A demon can only attack once per rest cycle (4 cycles) regardless of AP cost, because attacks require exhaust.
Correct rule: Attacks require exhausting the demon. Demons are only readied at the Rest Phase (every 4 cycles) unless an ability readies them. AP cost reduction saves resources but does not increase attack frequency. Murmur's -2 AP saves ~2 AP per rest cycle on one attack — meaningful but not game-warping standalone.
Unmoving Time is powerful recurring lane lockdown
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Called Gaap's Unmoving Time 'recurring lane lockdown' and ranked Gaap #3 strongest demon. In reality, Unmoving Time is a 1-cycle stall: you spend 2 AP + Gaap's exhaust, opponent's demons can't act that cycle, but they just activate normally next cycle. No permanent damage, no kills, no CP forced. Costs more than it gains.
Correct rule: Unmoving Time delays opponents by 1 cycle at the cost of 2 AP + exhaust. The opponent's demons were going to exhaust from acting anyway. Temporary action denial without follow-up kill potential is not meaningful value.
Speed determines when demons act or deploy
| Impact | Low |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Repeatedly cited Fast speed as 'early deployment,' 'acts before most demons,' and 'deploys early.' Cited Slow speed as making demons 'vulnerable.' Speed only affects Contract Phase priority (who chooses Main Phase order), not combat timing or deployment order.
Correct rule: Speed (Fast/Normal/Slow) only determines Contract Phase priority. Fast demons help you win the Contract Phase, letting you choose who takes their Main Phase first. It has no effect on combat timing, deployment order, or vulnerability.
Furfur Lie payoff is easily denied
| Impact | High |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Treated Furfur's Lie payoff (2 actions with -2 AP, ignoring exhaust/ready) as a near-guaranteed burst. Failed to reason through the Question familiar's Challenge mechanic and the fact that the opponent always has a freshly contracted demon each cycle to Challenge with.
Correct rule: Question's Challenge action is Universal, 0 AP, requires ready. The opponent contracts a new (readied) demon every cycle and can place it in Question's lane to Challenge. If Lie, payoff denied at near-zero cost. If Truth, only 3 Fixed Damage to local allies. The Lie payoff is conditional on the opponent not being able to or choosing not to Challenge — not a reliable burst tool.
CP framed as opponent's gain instead of your loss
| Impact | Low |
| Category | Terminology |
| Corrected | Yes |
What went wrong: Described VP/CP from demon death as points 'given to the opponent' or that the opponent 'earns'. In Evokers, when your demons die you accumulate CP toward your own loss — CP is a loss condition you suffer, not a reward your opponent collects.
Correct rule: When a demon dies, its controller gains CP equal to the demon's VP. Reaching 15 CP means you lose. CP should be framed as a cost you pay when your demons fall, not as something your opponent receives.
Judas Iscariot rated F tier — missed Exile exhaustion lock and solo lane play
| Impact | High |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Rated Judas F tier because the aura hurts your own board and the opponent can 'just move him away' or 'Exile him back'. Missed three critical interactions: (1) Exile leaves Judas exhausted, so the opponent cannot march him or re-Exile until a Rest Phase — that's multiple cycles of lane lockdown. (2) If Judas is your only demon, his aura does nothing negative since it affects 'Other' local allied demons. (3) Damaging Judas before Exile makes him a CP time bomb — if he dies on the opponent's side, they eat 3 CP.
Correct rule: Exhausted demons cannot perform any actions, including March. Exile exhausts Judas on transfer. The opponent has no cheap way to remove Judas from their lane until the next Rest Phase. Judas's aura targeting 'Other' local allied demons means a solo Judas has zero downside for the controller.
Thanatos + Focalor rated as a combo — two independent AoE nukes don't synergize
| Impact | Low |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed Thanatos + Focalor as a combo because Thanatos adds +3 fPWR to Focalor. But +3 PWR on a demon already using Focus (+5 PWR) is marginal, and both abilities are expensive exhaust-gated AoEs competing for the same AP and action slot. Two versions of the same effect on one fused demon is worse than two independent threats.
Correct rule: A real combo requires abilities that ENABLE each other (debuff + damage, positioning + AoE, AP discount + expensive ability), not two versions of the same effect stapled together. Fusion is only a combo if the bottom card's abilities meaningfully change what the top card can do, not just add a small stat bonus.
Judas + Gamigin listed as a combo despite no amplification
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Created a combo entry for Judas + Gamigin 'Double Sabotage' that was just two independent disruption effects with no interaction. Judas locks a lane via Exile exhaustion, Gamigin forces fusion on death — but neither ability makes the other stronger. This is two good demons played in the same game, not a combo. A real combo requires amplification: demon A makes demon B's ability do something it couldn't do alone, producing an effect greater than the sum of parts.
Correct rule: A combo requires synergy where abilities amplify each other — the combined effect must be greater than the sum of individual effects. Two demons doing unrelated things in parallel is a draft strategy, not a combo. Before logging a combo, ask: does demon A make demon B stronger (or vice versa)? If they just coexist, it's not a combo.
Ipos + Sekhmet listed as combo despite redundant effects
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Created a combo entry for Ipos + Sekhmet 'Mass Ready + Blitz' claiming Sekhmet's exhaust-free attacks + Ipos's mass ready was a synergy. But Sekhmet's Relentless Attack already removes the exhaust cost from attacks — if allies don't exhaust, Ipos's readying ability has nothing to ready. The two abilities overlap rather than amplify. At best Ipos covers demons that used non-Relentless actions, but that's marginal utility, not a combo.
Correct rule: Two abilities that solve the same problem (exhaust) in different ways are redundant, not synergistic. A combo requires abilities that address different bottlenecks or create effects neither could achieve alone. Exhaust bypass + readying = solving the same constraint twice.
Mammon analysis assumed opponent gets fresh Mammon on lost initiative
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Claimed losing initiative gives the opponent a readied Mammon. Contracted demons come into play during YOUR main phase, so you always get first use of Mammon regardless of who acts first. The opponent only gets a readied Mammon post-rest if they have initiative — not on the initial play.
Correct rule: Contracted demons enter play during your main phase, not during the contract phase or the opponent's main phase. On the turn Mammon is played, the controller always gets first use. Initiative only matters for post-rest Mammon control.
Claimed Slow demons 'die before doing anything'
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Stated that Gamigin's Slow speed and 6 HP means it 'dies before doing anything useful'. Speed does not affect when a demon enters play — all contracted demons come into play during the main phase. Speed only determines who chooses main phase order during the Contract Phase.
Correct rule: Contracted demons are played during your main phase regardless of speed. Slow speed means the opponent's Evoker likely chooses main phase order (acts first), but the Slow demon still enters play and can act during your main phase.
1x interpreted as once per game instead of once per turn
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Treated 1x actions as single-use abilities that can only be performed once ever. This led to undervaluing many demons whose 1x abilities were assessed as 'one-shot' effects. Multiple tier ratings were affected.
Correct rule: 1x means once per turn, not once per game. A demon with a 1x action can perform it every turn/cycle, just not multiple times within the same turn.
Judas + Gamigin combo write-up ambiguous about independent vs fused demons
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: The combo description said 'Stacks two independent disruption effects' without clarifying whether Judas and Gamigin are two separate (unfused) demons operating independently or a fused pair. A reader could interpret the combo as requiring a Judas-Gamigin fusion, which would be a completely different setup with different AP costs and positioning.
Correct rule: Judas + Gamigin Double Sabotage uses two independent, unfused demons in separate lanes. Each creates its own disruption: Judas via Exile exhaustion lock, Gamigin via mandatory death-trigger fusion onto an opponent's demon. The combo's strength is parallel disruption across two fronts — fusing them together would defeat the purpose.
Naberius + Buer listed as combo despite no amplification
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Created a combo entry for Naberius + Buer 'Resilient Frontline' claiming healing + death trigger was a synergy. Buer heals local allies, Naberius has a death trigger for a free action. Buer just delays Naberius dying — that's sustain next to a body, not amplification. Buer doesn't make the death trigger stronger, Naberius doesn't make healing better. Two decent demons in the same lane is not a combo.
Correct rule: A combo requires amplification: demon A must make demon B's ability do something it couldn't do alone, producing an effect greater than the sum of parts. Sustain + death trigger is not amplification — the death trigger fires the same way regardless of whether Buer is present. Healing just delays the timeline.
Vine + Libra listed as combo despite no amplification
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Created a combo entry for Vine + Libra 'CP-Free Trades' claiming Libra's mutual kill + Vine's death spawns was a synergy. Libra kills one demon from each side with CP ignored. Vine spawns Crystal Parasites from 3+ CP deaths. But Libra doesn't make Vine's spawns stronger, and Vine doesn't make Libra's kills more effective. They just coexist — Libra creates deaths, Vine reacts to deaths, but neither amplifies the other.
Correct rule: Two demons that independently react to the same game event (deaths) without making each other's abilities stronger are not a combo. A combo requires demon A to amplify demon B's effect beyond what B could achieve alone.
Great Blue Spot described as enemy-only AoE
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Repeatedly described Focalor's Great Blue Spot as dealing damage 'across a lane' or 'to a lane' without noting it hits ALL demons including your own allies. The ability text says 'All Demons in Target Lane' not 'All Enemy Demons'. This means friendly fire is a real cost — allies in the target lane take equal damage and their deaths give you CP.
Correct rule: Great Blue Spot divides damage between ALL Demons in the target lane, including allied demons. To avoid friendly fire, target a lane with zero allies. The opponent can counter by positioning a demon in a lane with your allies, forcing you to choose between friendly fire or not using the ability on that lane.
Quick actions resolve AFTER the triggering action — wrong, they resolve BEFORE
| Impact | High |
| Category | Timing |
| Corrected | Yes |
What went wrong: Assumed Quick actions are reactive responses that fire AFTER the opponent's action resolves. Concluded that Sabnock's Pink Haze (-3 PWR) could not prevent damage from an incoming 2 PWR attack because 'the damage already happened'. In reality, Quick actions resolve BEFORE the triggering action — the opponent declares an action, the Quick response resolves first, then the declared action resolves with any modifications from the Quick response.
Correct rule: Quick actions resolve BEFORE the action they are responding to. The opponent declares an action, the defending player may respond with 1 Quick action, the Quick resolves first, then the original declared action resolves. This means a Quick debuff (like Pink Haze -3 PWR) applied to a 2 PWR attacker reduces their effective PWR to 0 before the attack deals damage. Sabnock can prevent all damage from a 2 PWR attack by responding with Pink Haze.
Sitri + Malphas listed as combo — Sitri's pull requires demons already exhausted, Malphas exhaust is redundant
| Impact | Low |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed Sitri + Malphas as a 'Lane Control' combo: Sitri pulls all exhausted demons to Sitri's lane at end of phase, Malphas exhausts enemies entering his lane. But Sitri's pull only moves ALREADY EXHAUSTED demons. If a demon is already exhausted, it can't act anyway (confusion #4). Malphas exhausting them after they arrive is redundant — they were already exhausted before being pulled. Malphas doesn't amplify Sitri's pull because the pulled demons are already locked out of actions. The two abilities solve the same problem (denying enemy actions) without amplifying each other.
Correct rule: When evaluating combos involving exhaust mechanics, check: are the targets already exhausted before the second ability fires? If yes, the second exhaust is redundant. Sitri pulls exhausted demons (already can't act), Malphas exhausts demons that enter (already exhausted from Sitri's requirement). Two exhaustion sources on already-exhausted targets is not amplification.
Nahash Forbidden Fruit AP reduction applied to wrong target
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Implemented Forbidden Fruit as applying -2 AP Cost to the MARKED DEMON's own actions (status 'ap_cost' on target). The card text says 'Actions that target the targeted Demon have -2 AP Cost' — the reduction is on ANY action TARGETING the marked demon, not the marked demon's own actions. The wrong implementation made Nahash look like it was helping the enemy (reducing their AP costs). The correct version is a team force multiplier: all allies can attack the marked target for 0 AP (2 base - 2 target modifier).
Correct rule: Forbidden Fruit's -2 AP Cost applies to ACTIONS TARGETING the marked demon, not to the marked demon's own actions. Any demon attacking the marked target pays 2 less AP. This is stored as 'ap_cost_when_targeted' status on the target, checked by execute_ability and game_loop attack cost calculations. This mechanic enabled the Nahash + Kimaris combo and warranted a tier increase from C to B.
Malthus + Capricorn Musket Fire Volley listed as combo — misunderstood Collective Unconscious limitations
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed Malthus + Capricorn as a burst combo where multiple allies chain Collective Unconscious to copy Musket Fire repeatedly. Two errors: (1) Musket Fire is already an Allied action — any allied demon can perform it directly without needing Capricorn to copy it. CC adds nothing that Allied doesn't already provide. (2) Collective Unconscious applies a Status on the performer that prevents CC from being performed again. The status is global — once any demon uses CC, that demon is locked out. The 'chain' of N allies each using CC doesn't work because CC's status prevents repeated use. The combo is just 'everyone uses Musket Fire' which is already how Allied actions work.
Correct rule: Allied actions are already performable by any allied demon (game-mechanics.md). Capricorn's Collective Unconscious copies the LAST action, but its status locks the performer from using CC again. When evaluating copy-action combos, check: (1) is the source action already Allied? If so, copying adds nothing. (2) Does the copy mechanism have a use-limiting status? If so, chains don't work.
Shax + Thanatos Kill Chain listed as combo — friendly fire kills own board on repeated use
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed Shax + Thanatos as an AP-positive kill chain engine: Styx's Shores hits ALL Other Demons for 7 PWR, kills generate AP + ready Shax, enabling repeated Styx uses. Missed that 'All Other Demons' includes your OWN demons (confusion #21). The second Styx use hits your own weakened allies too. After two Styx uses, every demon on the board (yours and theirs) has taken 14 PWR damage. Your own board is wiped alongside the enemy's. This is not an engine — it's a murder-suicide that scales equally against both sides.
Correct rule: AoE abilities that say 'All Other Demons' hit your own allies (confusion #21, game-mechanics.md). When evaluating AoE engine combos, always calculate the FRIENDLY FIRE cost across multiple uses. Shax+Thanatos dealing 7 to all demons twice = 14 total to your own allies. Any combo that requires repeated AoE self-harm is not viable unless your own demons can survive (e.g., high DEF or healing to offset).
Mammon + Purson combo analysis invented a CP bomb mechanic that doesn't exist — Purson only has 1 familiar out at a time
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: When asked about Mammon + Purson, invented an elaborate 'CP bomb delivery' combo where Mammon transfers control of 4 Purson familiars to the opponent, then kills them for 8 CP. Multiple errors: (1) Purson's Bestow Familiar explicitly says 'If none of Purson's Familiars are in play, play 1' — only ONE familiar can be out at a time. Not 4. (2) Built an entire combo theory around a mechanic that doesn't exist (mass familiar transfer as CP payload). (3) Got excited by the 'fused demons have both names' insight and ran with it without checking the actual constraints on Purson's familiar deployment. This is the pattern of confusing 'sounds cool' with 'actually works' that has plagued the combo-finder.
Correct rule: Always read the FULL ability text including constraints. Purson says 'If none of Purson's Familiars are in play' — this limits deployment to 1 familiar at a time. When building combo theories, check every constraint on every ability involved before elaborating. Don't build multi-step theories on top of an unchecked assumption.
Vine + Libra listed as combo — opponent controls Libra's targeting and can kill Vine first
| Impact | Low |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed Vine + Libra as a 'CP-Free Trades' combo: Libra's Zero Sum kills one demon from each side (CP ignored), Vine spawns Crystal Parasites from 3+ CP deaths. Missed that Zero Sum says 'you and your opponent EACH Target Any Demon' — the opponent chooses what dies on YOUR side. The opponent simply targets Vine with their pick, killing Vine before any Parasites can spawn. Even if Vine survives the first Zero Sum, the opponent can choose to kill Vine on any subsequent use. The combo is self-defeating: the mechanic that's supposed to enable it (Zero Sum) gives the opponent the tool to disable it.
Correct rule: When evaluating combos involving opponent-choice mechanics (like Zero Sum where both players pick targets), check if the opponent can use their choice to disable the combo. If the opponent can simply target the enabler card, the combo doesn't work against a competent opponent.
Castor/Pollux 'Other Demon' exclusion applies to BOTH, not just the projector
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Implemented Azazel's aura 'All Other Local Allied Demons' projected via Castor/Pollux sharing with standard self-exclusion (only the projecting demon is excluded). Concluded Castor gets +2 DEF from Pollux's shared copy and vice versa. The card text explicitly states: 'When Castor or Pollux resolves or applies an ability, if the text uses the keyword Other Demon(s), it excludes BOTH Castor and Pollux.' So neither Castor nor Pollux benefits from any 'Other' ability projected by either of them.
Correct rule: When any ability with 'Other Demon(s)' in its text is resolved or applied by Castor OR Pollux (including shared abilities from fused cards), BOTH Castor and Pollux are excluded from 'Other'. This means: Castor-Azazel aura buffs Sabnock but NOT Pollux. Pollux's shared Azazel aura buffs Sabnock but NOT Castor. Neither twin ever benefits from the other's 'Other' abilities.
Zagan Omnipresent Striker listed as combo — it's just how Zagan works solo
| Impact | Low |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed 'Zagan Omnipresent Striker' (Zagan fused under a high-PWR demon for reactive repositioning) as a combo. But Zagan's entire passive IS reactive repositioning when fused — 'After an action is declared, move Zagan to Any Lane.' Fusing Zagan under any strong demon is literally the only way to use Zagan. It works with ANY demon, not a specific pairing. This belongs in a Zagan character guide, not the combo section.
Correct rule: If the interaction works with ANY partner demon and is the card's core designed use case, it's a character guide topic, not a specific combo. Zagan under 'a high-PWR demon' is generic advice, not a two-card synergy.
Andras + Lix Tetrax listed as combo — no amplification, AP efficiency is generic
| Impact | Low |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed Andras + Lix Tetrax as an 'AP Efficiency' combo: Lix Tetrax's passive applies -1 AP Cost to targets, making Andras's Momento Mori cheaper. But -1 AP Cost is a generic benefit that helps ANY demon's abilities, not specifically Andras. Lix Tetrax doesn't make Andras's lock+attack DO something it couldn't do alone — it just makes it slightly cheaper. 'Slightly cheaper' is not amplification, it's incremental improvement. This is the same as saying 'Murmur + anyone is a combo because -2 AP Cost.' Generic AP reduction paired with any ability is not a specific combo.
Correct rule: AP cost reduction paired with any ability is generic synergy, not a specific combo. If the AP reducer works equally well with any other demon, it's not a two-card combo — it's a character guide topic for the AP reducer. The combo must produce a qualitatively different effect, not just a quantitatively cheaper one.
Lucifer + Zepar listed as combo — no amplification, just two redirect passives coexisting
| Impact | Low |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed Lucifer + Zepar as a 'Forced AoE Splash' combo where Lucifer reflects damage to one enemy and Zepar redistributes it across all enemies. But there's no amplification: Lucifer takes X damage and reflects X-2 to one target. Zepar just splits that X-2 across multiple targets instead of one. The total damage output is still X-2 regardless of Zepar. Zepar doesn't make Lucifer's splash STRONGER — it just changes the distribution. Two redirect passives coexisting in the same lane is not a combo, it's two demons doing independent things.
Correct rule: A combo requires amplification: demon A must make demon B's ability do something it couldn't do alone, producing an effect greater than the sum of parts. Redistributing the same total damage across more targets is not amplification — it's the same total damage with a different spread. Check: does the combination produce MORE total output than either demon alone? If not, it's coexistence, not a combo.
Ipos + Sekhmet listed as combo — redundant exhaust bypass, heavily AP-gated
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed Ipos + Sekhmet as a 'Mass Ready + Blitz' combo: Sekhmet's Relentless Attack lets allies attack without exhausting, Ipos readies exhausted demons for a second wave. Two problems: (1) Sekhmet's Relentless Attack already ignores exhaust — allies attack WITHOUT exhausting. If allies don't exhaust, Ipos has nothing to ready. The two abilities solve the same bottleneck (exhaust) redundantly. (2) Both abilities are heavily AP-gated: Relentless Attack costs 2 AP per use, Ipos Command Army costs 1 AP. A 'blitz' of 3 Relentless Attacks + Command Army = 7 AP, far exceeding the 3 AP per contract phase. The combo requires banking AP across multiple phases for a single burst turn that doesn't justify the investment.
Correct rule: Two abilities that bypass the same bottleneck (exhaust) are redundant, not synergistic. Also verify the total AP cost of the 'combo turn' — if it exceeds available AP by a large margin, the combo is impractical. Sekhmet's value is that allies DON'T exhaust, making readying pointless. AP is the real constraint (confusion #12).
Morpheus + Nergal listed as combo — Nergal's cannot-be-readied only affects allied demons, not enemies
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed Morpheus + Nergal as a 'Sleep Lock' combo: Morpheus exhausts all local demons at end of phase, Nergal prevents enemies from readying. But Nergal's abilities only affect YOUR demons: Adaptability targets 'Target Demon' (Allied action — performed by allies on allies), Nine Variations applies 'All Allied Demons have -1 AP Cost and cannot be Readied.' Neither ability can target or affect enemy demons. The combo's premise (Nergal locking enemies from readying after Morpheus exhausts them) is impossible — Nergal has no way to apply cannot-be-readied to enemies.
Correct rule: When evaluating combos, verify the TARGETING of each ability. Allied actions target allied demons only. 'All Allied Demons' means YOUR demons, not the opponent's. Nergal's 'cannot be Readied' status is a self-imposed downside for the AP cost reduction, not an offensive tool. Always read the targeting text: Allied = your side only.
Dantalion Defensive Fortress listed as combo — it's just how Dantalion works solo
| Impact | Low |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed 'Dantalion Defensive Fortress' (Dantalion + Duban + Taurus + Azazel + Zepar) as a combo. But stacking defensive demons onto Dantalion is literally Dantalion's core mechanic — 'Any number of Demons can be Fused onto Dantalion.' Fusing defensive passives onto an unlimited-fusion demon is not a combo between two cards, it's one card functioning as designed. The 'combo' is just a list of good fusion targets for Dantalion, which belongs in Dantalion's character guide (where it already is), not the combo section.
Correct rule: A combo requires two specific cards whose abilities amplify each other in a non-obvious way. A demon using its core mechanic (unlimited fusion) with generically good fusion targets is not a combo — it's a character guide topic. If the interaction works with ANY defensive demon (not specifically these four), it's not a specific combo.
Ronove + Auns combo rated C+ tier — missed Willpower exhaust bypass making it an 8x damage engine
| Impact | High |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Rated the Ronove + Auns fusion as C+ tier because 'exhaust is the real bottleneck, not AP' and 'Murmur does the AP job better.' Completely missed that Auns has Willpower (0 AP, 1x): Ready self + 3 Fixed Damage to self. This breaks the exhaust bottleneck that was cited as the limiter. The actual cycle is: Attack (exhaust, 0 AP) → Willpower (ready) → Attack again = 2 attacks per turn for 0 AP. Over 4 cycles that's 32 damage from a single demon vs 4 from a normal demon. 8x output.
Correct rule: When analyzing fusion combos, check ALL abilities on BOTH cards, not just the passives. Auns has two abilities: the AP-to-damage passive AND Willpower (ready + self-damage). The passive alone is mediocre (confusion #12: exhaust bottleneck). But Willpower bypasses exhaust, which combined with Ronove's heal creates a self-sustaining damage engine. Always check if the bottom card has an active ability that synergizes with the top card's passive.
Nahash + Kimaris listed as combo — chaining a normal action with a Start of Turn action across different phases
| Impact | High |
| Category | Timing |
| Corrected | Yes |
What went wrong: Listed Nahash + Kimaris as a 'Lethal Setup' combo: Nahash uses Forbidden Fruit (normal action during your main phase) to debuff a target with -2 DEF, then Kimaris uses The Evening Bell Tolls (Start of Turn — fires at start of OPPONENT'S main phase) to execute targets at exact HP thresholds. Critical timing error: Nahash's Forbidden Fruit is a normal action performed during YOUR main phase. Kimaris's Evening Bell Tolls is Start of Turn, which fires at the start of the OPPONENT'S main phase. These are different phases. Nahash's -2 DEF status expires at end of YOUR main phase (status effects are temporary). By the time Kimaris's Start of Turn fires next phase, the -2 DEF is already gone. The combo's premise — Nahash softening targets for Kimaris's execute — requires effects to persist across phases, which they don't.
Correct rule: Status effects expire at end of current main phase (game-mechanics.md). Start of Turn abilities fire at the start of the OPPONENT'S main phase, which is a DIFFERENT phase from when normal actions resolve. Any combo chaining a normal action's status with a Start of Turn ability fails because the status expires before the SoT fires. Always check: are both abilities in the SAME phase? If not, status effects from the first won't be active for the second.
Naberius + Buer listed as combo — healing doesn't amplify Naberius's death trigger
| Impact | Low |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed Naberius + Buer as a 'Resilient Frontline' combo claiming Buer's healing keeps Naberius alive and guarantees the death strike. But Naberius's key ability is a death trigger (free action when fatally wounded). Healing DELAYS the death trigger — it doesn't amplify it. Buer keeping Naberius alive means the death trigger fires LATER, not BETTER. The combo fails the amplification test: Buer doesn't make Naberius's death strike stronger, and Naberius doesn't make Buer's healing stronger. They just coexist — healer next to a demon with a death trigger.
Correct rule: Sustain + death trigger is not a combo (already listed in the NOT combos section). Healing delays death but doesn't amplify the death trigger's effect. For a real combo with Naberius, find something that makes the free action on death MORE powerful (e.g., a PWR buff that makes the parting strike lethal), not something that delays when it fires.
Judas + Gamigin Double Sabotage listed as combo — no amplification, two independent disruptions
| Impact | Low |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Listed Judas + Gamigin as a combo despite this being flagged as invalid earlier in the session (confusion logged at the start). Judas locks a lane via Exile exhaustion, Gamigin forces fusion on death. Neither ability makes the other stronger. This was already identified as 'two good demons played in the same game, not a combo' but the combo entry was never removed from Notion. A real combo requires amplification: demon A makes demon B's ability do something it couldn't do alone.
Correct rule: Two independent disruption effects operating in parallel are a draft strategy, not a combo. This was already logged as a confusion earlier. The combo should have been removed when the original confusion was logged.
Auns AP-to-damage passive implemented as no-op marker instead of actual mechanic
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Implemented Auns (#042) passive 'When this would spend X AP, deal X Fixed Damage to Auns instead' as a no-op marker passive with an unused helper function. The marker was never wired into execute_ability, so the AP-to-damage conversion never actually happened. Auns just spent AP normally like any other demon. This also meant interactions with other abilities (like Ronove's lifesteal triggering on the self-damage) never worked.
Correct rule: Auns passive must intercept AP deduction in execute_ability. Before spending AP: check if Auns ability is active (unit_id '042' or fused_bottom '042'), check remaining HP >= cost, then deal_fixed_damage to self instead of deducting AP. The self-damage must fire a DAMAGE_DEALT event so triggered passives like Ronove's lifesteal can respond. When Ronove is fused with Auns, the self-damage triggers Ronove's heal, creating a sustain engine.
Sekhmet Quick suppression blocked all ACTION_DECLARED triggers
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Implemented Sekhmet's passive ('Quick actions cannot be performed in response to Relentless Attack or Relentless March') by suppressing the entire ACTION_DECLARED event. This incorrectly blocked passive triggers like Zagan's reposition, which is a passive ability that fires on ACTION_DECLARED — not a Quick action. The suppression should only prevent Quick-timed actions from being performed in the response window, not prevent passive triggers from firing.
Correct rule: Sekhmet's passive only suppresses Quick actions (the timing type) in response to Relentless abilities. Passive triggers that fire on ACTION_DECLARED (like Zagan's reposition, Orias's extra target) are NOT Quick actions and must still fire normally. ACTION_DECLARED serves two purposes: (1) fire passive triggers, (2) open the Quick response window. Sekhmet blocks #2 only.
Ose Insanity implemented as status buffs instead of bonus action grant
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Implemented Ose's Insanity ('The targeted Demon's Evoker may have the targeted Demon perform an action with -2 AP Cost and ignoring exhaust/ready costs') by only applying discount statuses (ap_cost -2, ignore_exhaust, ignore_ready) to the target. Did not actually grant the bonus action — just assumed the normal turn flow would pick up the statuses. This is wrong: the card explicitly says 'may have the targeted Demon perform an action', which is an immediate bonus action grant during Insanity's resolution, not a buff for later. Additionally, the target can be an enemy demon, meaning the enemy player controls the bonus action — the statuses alone don't give the opponent a turn.
Correct rule: Ose's Insanity grants an immediate bonus action to the target's controller. The target performs one action right now with -2 AP Cost and ignoring exhaust/ready costs. This requires: (1) temporarily setting current_player to the target's owner so timing checks pass, (2) readying exhausted targets since ignore_exhaust is active, (3) executing the action inline during Insanity's resolution. The same pattern applies to any card that says 'may have [demon] perform an action' — it's an immediate grant, not a deferred buff.
Push-away abilities inconsistently handled same-lane and boundary cases
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Three demons have 'move X Lanes Away From [Demon]' mechanics: Sharur (#068), Scorpio (#062), and Amun (#086). Claude implemented each independently at different times and got different things wrong:
- Sharur: same-lane targets returned None (no push). Treated 'away from' as requiring a directional vector.
- Amun: same-lane at max boundary (both lane 2) computed min(2, 2+1) = 2 (no movement). Missed that 'away' at a boundary means the other direction.
- Scorpio: implemented correctly (happened to have a boundary fallback).
The root cause is treating each push ability as a unique snowflake instead of recognizing the shared pattern. 'X Lanes Away From [Demon]' is a single mechanic that appears on 3 cards. It should have been implemented as a shared helper function with consistent rules, not 3 separate implementations with 3 separate bugs.
Correct rule: 'Move X Lanes Away From [Demon]' always results in the target ending up further from the source demon. The rules:
- Target in different lane: push further from source, clamped to board edges [0,2].
- Target in same lane: push to adjacent lane (out of source's lane). Prefer higher lane number; if at max boundary (lane 2), push to lower.
- There is NO case where the push doesn't happen (except source fatally wounded).
- This is a shared mechanic — all 'Away From' abilities follow the same rules. Should be a single helper function, not duplicated logic.
Sharur same-lane push treated as no-push
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Implemented Sharur's passive ('move targeted Demon 1 Lane Away From Sharur') with a same-lane case that returned None (no push). Interpreted 'away from' as requiring a directional vector — when source and target are co-located, there's no vector, so no push. This is wrong: 'away from Sharur' means the target must end up NOT in Sharur's lane. Same-lane targets should be pushed to an adjacent lane.
Correct rule: '1 Lane Away From Sharur' means the target must always end up in a different lane from Sharur after the push. If target is in the same lane, push to an adjacent lane (default: higher lane number, or lower if at max boundary). There is no 'no push' case — the only time push doesn't happen is if Sharur is fatally wounded.
Sitri bonus action not implemented despite Vual having the same mechanic
| Impact | High |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Sitri's Siren's Call and Vual's Sun Does Not Set both have end-of-phase bonus actions ('may perform an action, ignoring exhaust costs'). When implementing Vual, Claude built the full bonus action pipeline: sun_bonus_action status checked in end_main_phase step 0, pending_bonus_actions on GameState, game loop granting the action. But Sitri was implemented in an earlier batch with a TODO: 'The end-of-phase free action for Sitri is out of scope for this batch.' When the infrastructure was built for Vual, Claude never went back to wire Sitri into the same system. Additionally, Sitri's movement effect (move all exhausted demons to Sitri's lane) has a helper function but is never called from end_main_phase.
Correct rule: End-of-phase bonus actions are a shared mechanic. Any card that says 'at the end of this Main Phase, [demon] may perform an action' should use the same pipeline: (1) apply a status marker during the ability, (2) check the marker in end_main_phase before status expiry, (3) add to pending_bonus_actions, (4) game loop grants the action. Sitri needs both the movement effect AND the bonus action wired into end_main_phase, using the same pattern as Vual.
Paimon Omnipotence had no timing gate on borrowed abilities
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Paimon's Omnipotence ('Reveal a Fusible demon card from your hand. Paimon performs a Special Action without Familiar from that demon card') was implemented as a simple handler lookup: find the borrowed ability by (unit_id, idx), call it. No filtering. This meant Paimon could borrow Quick abilities (like Pink Haze), Start of Turn abilities (like Vual's Sun Does Not Set), Gear abilities, passives, and even familiar abilities — all of which the card explicitly blocks.
The reminder text on the card literally says: 'As Omnipotence does not have Quick or Start of Turn, Special Actions with Quick or Start of Turn cannot be performed in their special phases.' And the main text says 'without Familiar.' Both constraints were ignored.
Root cause: the ability-writer agent implemented Omnipotence as a generic 'call any handler' function without reading the reminder text (idx=1) on the same card. The reminder is a rules clarification, not flavor text — it defines hard constraints. The agent pattern-matched 'borrow another demon's ability' as 'look up handler and call it' without considering what kinds of abilities are excluded.
Correct rule: Paimon's Omnipotence can only borrow abilities with Action, Allied, or Universal timing from non-familiar demon cards. Blocked timings: Quick, Gear, Start of Turn, Passive. Blocked sources: familiar cards. The gate must check get_ability_timing() on the borrowed ability before executing. Reminder text on cards is mechanically binding — always read ALL ability indices on a card, not just the one being implemented.
FATALLY_WOUNDED event doesn't fire from trigger-chain damage
| Impact | High |
| Category | Interactions |
| Corrected | Yes |
What went wrong: When Stolas's trigger deals 1 damage via deal_fixed_damage inside an event handler, and that damage kills a demon (sets fatally_wounded=True), no FATALLY_WOUNDED event fires. This is because FATALLY_WOUNDED events are only fired in execute_ability step 10 — which only runs after ability handler resolution. But Stolas's trigger runs inside fire_event (the event dispatch loop), not inside execute_ability. So deal_fixed_damage sets the flag, but nothing fires the FATALLY_WOUNDED event for other triggers (like Camio's Black Serpent deployment) to respond to.
This means any kill that happens inside an event handler chain (trigger damage killing a demon) does NOT fire FATALLY_WOUNDED. Camio's 'When Camio Fatally Wounds a Demon' trigger would never see these kills.
Root cause: FATALLY_WOUNDED event firing was only added to execute_ability (the ability execution path), not to the general damage functions. deal_fixed_damage and deal_damage set the fatally_wounded flag but don't fire the event — they rely on the caller to do it. When the caller is fire_event (not execute_ability), nobody fires the event.
Correct rule: deal_fixed_damage and deal_damage should fire FATALLY_WOUNDED events when they set the fatally_wounded flag, OR fire_event should check for newly fatally wounded demons after each trigger handler returns and fire FATALLY_WOUNDED for them. The latter approach is more robust since it catches all trigger-chain kills regardless of the damage source. Currently the flag is set but the event is missing from the trigger chain path.
Fused bottom card triggers never fired
| Impact | High |
| Category | Interactions |
| Corrected | Yes |
What went wrong: The event system's get_triggered_abilities() in events.py built a lookup of unit_id -> demons on the field, but only indexed by demon.unit_id (the top card). When Camio (#098) is fused with Stolas (#022), the fused demon's unit_id is '098'. Stolas's DEMON_EXHAUSTED trigger is registered under unit_id '022'. The lookup never found '022' on the field because no demon has that as its top card — the fused demon is indexed as '098'.
This means ALL fused bottom card triggers were silently broken. Not just Stolas — every bottom card with a register_trigger() call (Seir death triggers, Belphegor deploy triggers, Dantalion death-fuse, etc.) would fail to fire when fused under another demon.
Root cause: the event system was built before fusion was implemented. When fusion was added, the passive system was updated (_on_field checks fused_bottom), but the trigger system was never updated to match. Two systems that should follow the same pattern diverged because they were implemented at different times.
Correct rule: get_triggered_abilities() must index demons by BOTH unit_id (top card) AND all fused_bottom IDs (comma-separated). When a trigger is registered for unit '022' and a demon on the field has fused_bottom='022', that demon should match. The fix: iterate fused_bottom.split(',') and add each to the unit_id->demons lookup. This mirrors how _on_field already checks fused_bottom for passives.
Misread 'In Range' as a generic proximity term instead of a fusion top-card Range reference
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: When reviewing a patch note that changed Andromalius's Quick Draw from 'In Range' to 'Distant,' Claude interpreted 'In Range' as a flexible/generic targeting term meaning 'nearby' or 'within melee reach,' and framed the change as 'brawler-sniper to pure backline reach.' This treated 'In Range' as a role descriptor rather than a defined mechanical keyword.
Correct rule: 'In Range' is a specific Evokers mechanic that references the TOP CARD's Range stat in a fusion context — it means 'within the range value currently supplied by the top card,' not a generic adjacency or flexibility term. Changing an ability from 'In Range' to 'Distant' moves it from fusion-top-range-dependent (variable, depends on the top card) to a fixed Distant range. Every capitalized range word in ability text (In Range, Distant, Local, Any, etc.) is a defined keyword with specific mechanical meaning — not an English-language descriptor.
Undervalued old Vapula's damage-to-heal passive, missing that it was the deliberate enabler of Cycle of Life's selective heal/damage split
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: When comparing old Vapula's kit (Cycle of Life: lockdown + AoE damage to All Other Local Demons [friendly fire] + damage-to-heal conversion passive) to new Vapula's kit (march tax passive + Cycle of Life: heal Allies + damage Enemies), Claude concluded new Vapula was stronger. This was wrong. Old Vapula's passive ('Whenever Vapula would deal X damage to Any Other Demon, you may remove X damage instead') was specifically designed to convert friendly fire from Cycle of Life into per-target healing. The net swing of old Cycle of Life WAS heal-allies-damage-enemies in optimal play — the new version just bakes that split into the explicit ability text. Old Vapula had the SAME heal/damage swing PLUS lockdown on top.
Correct rule: When evaluating a card whose ability creates friendly fire, always check whether the demon has passives that interact with self-dealt damage, and work out whether the passive converts, prevents, or mitigates the friendly-fire portion. A 'convert damage to heal' passive on Any Other Demon, paired with an ability that deals damage to All Other Local Demons, is a deliberately coupled design: the AoE plus conversion passive IS a heal-allies-damage-enemies ability by design. Do not treat the passive as 'niche' or 'rarely used' without tracing its intended interaction with the demon's own abilities.
General principle: when an old version of a card gets simplified to make an effect explicit, the simplification often doesn't ADD power — it just reveals power that was already there via complex interactions. When judging 'is the new version stronger,' account for what the old version could ALREADY do in optimal play, not just what its text literally read.
Recommended Tyr Gleipnir for In Range, missing that Dromi exists to pull Distant targets INTO Local range for Gleipnir
| Impact | High |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Claude recommended converting Tyr's Gleipnir (Local exhaust) to In Range, citing 'Tyr's other ability Dromi is already Distant — kit mixes ranges.' This misread Tyr's kit. Dromi is a 0 AP pull that moves a Distant Demon to Tyr's Lane. The kit's intended flow: Dromi pulls target into Local range, then Gleipnir exhausts. Converting Gleipnir to In Range collapses the combo into a single-ability solution, invalidating Dromi's design purpose.
Correct rule: When a demon has multiple abilities, they are usually designed as a combo — each ability's range constraint often exists to make another ability meaningful. Removing a range constraint on the payoff ability can invalidate the setup ability.
Recommended Sharur Heavy Impact for In Range without reading the passive's push-away interaction
| Impact | High |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Claude recommended converting Sharur's Heavy Impact (Local damage) to In Range, calling the push-away passive 'synergy' for extended reach. Sharur's passive 'moves the targeted Demon 1 Lane Away From Sharur' after any action — the push exists specifically to knock Local-range hits into non-Local positions. If Heavy Impact targets a Distant demon via In Range fusion, the push becomes meaningless displacement; the ability becomes non-functional in its intended combo role.
Correct rule: Before recommending a range-keyword change on any ability, read ALL of the demon's passives and related abilities. Sharur's passive push is the whole point of his Local-targeting damage — it's a positional-displacement combo, not a standalone attack. Extending the range breaks the combo's purpose.
Claimed ambiguity over 'who benefits' from Gamigin's Fixed Damage passive, missing the default rule that the demon's controller controls its abilities
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: When explaining Gamigin's new passive 'Gamigin may deal damage as Fixed Damage,' Claude flagged it as ambiguous: 'could apply while Gamigin is alive on the original owner's side, OR apply to the fusion's damage after Gamigin is fused onto the opponent, OR both. I don't know which.' This was a false uncertainty — there is no ambiguity. In Evokers, the controller of a demon controls that demon's abilities and special actions, and that principle resolves the question directly: whoever currently controls Gamigin (or the fusion containing Gamigin) is the one who chooses when to invoke the Fixed Damage option.
Correct rule: Default control rule: the controller of a demon controls that demon's abilities and special actions. This applies universally unless an ability's text explicitly overrides it (e.g. ua abilities which let any performer invoke them, or text clauses like 'If you don't control [Demon]' which gate the effect). For Gamigin's Fixed Damage passive, no such override is printed — so it defaults to: whoever currently controls Gamigin controls the Fixed Damage option. Before Gamigin dies, that's the original owner. After death and forced-fuse onto the opponent's demon, that's the opponent. The 'passive follows the card' property falls out of the controller-controls-abilities rule directly.
Reasoned about demons from names/themes instead of reading card text
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: When evaluating which single-target Local demons would be interesting to convert to In Range, Claude split them into 'caster/commander archetypes (Andras, Berith, Flauros)' vs 'warrior/brawler ones (Marchosias, Sharur, Tyr)' based entirely on the demon names and mythological themes, without reading any of their actual ability text from the CSV. Claude recommended conversion candidates based on thematic guesses rather than mechanical reality.
Correct rule: When reasoning about any demon in Evokers, ALWAYS read the actual card ability text from the unit CSV (or the Abilities database / Rulesbook.pdf) before making design claims. Demon names and thematic archetypes do not determine their mechanics — the card text does. The rulebook (Rulesbook.pdf in EvokersState/) and the complete ability text in the Draft Design CSV are the authoritative sources; there is no reason to guess from a name when the text is directly available.
Framed 'In Range' as fusion flexibility when it's actually a fusion constraint
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: When analyzing Andromalius / Gusion / Nyx losing the 'In Range' keyword, Claude framed the change as 'removes a fusion-build lever' / 'removes a source of novel combos' / 'simplifies the fusion-build puzzle.' Claude treated In Range as if it were a source of flexibility — where players could stack the demon under different tops to modulate reach. This was backward.
Correct rule: 'In Range' is not a flexibility lever — it is a fusion-range CONSTRAINT. An In Range ability is locked to whatever range the current top card supplies. If you fuse the demon under a Local top, the In Range ability becomes Local-only, meaning you LOSE reach. Converting In Range → Distant (or any fixed range) actually INCREASES fusion flexibility, because the ability retains its reach regardless of what top is stacked. You can fuse under a Local top without giving up the ability's range. So the Andromalius/Gusion/Nyx conversions are fusion-flexibility BUFFS — not simplifications or nerfs to build variety. The designer is freeing these demons from a fusion penalty, not removing a build lever.
Framed SS on ua abilities as a meaningfully 'different flavor' of nerf, missing that ua is never quick — so the whole-board lockout never fires reactively
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: After learning ua means Universal Allied (both sides can perform it), Claude framed the patch's new SS locks on Dionysus's Drunken Match and Seir's Death Is the Great Equalizer as a 'whole-board lockout' that represented a 'different flavor of nerf' compared to SS on a regular Action. This overstated the interactivity. In Evokers, abilities are performed during the performer's main phase (not reactively, unless the ability is quick). There are no quick ua abilities in the game, so an opponent cannot 'steal' a ua ability's SS lockout mid-opponent-turn — they would only be able to use it on their own main phase, which is a separate window anyway. The functional effect of SS on ua ends up nearly identical to SS on a regular Action.
Correct rule: The ua → whole-board lockout property is technically correct (the SS applies to the ability itself, so once anyone triggers it, it's locked for the round for everyone). But this only matters when the opponent could have USED the ability in a way that interacts with yours — which requires the ability to be playable reactively. Abilities only resolve reactively if they have the quick keyword, and there are no quick ua abilities. So in practice, SS on a ua caps the ability at once per round just like SS on a regular Action: the opponent's separate main phase is a separate window anyway, so 'locking them out' doesn't change game timing in any meaningful way.
General principle: before describing a mechanic as a 'different flavor' of effect, check whether the difference is actually observable during gameplay given the other rules in the system. Cross-side access only matters when it creates a timing interaction (reactive usage during the opponent's turn). Without quick, there is no such interaction.
Recommended Morax Nurture (a Quick allied heal) for In Range conversion — same allied-target issue as Alchemy, but compounded by Quick creating opaque reactive counterplay at fusion-extended range
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Claude recommended converting Morax's Nurture (1 AP, quick, 1x/round, heal tpwr damage from Target Other Local Demon) to In Range, calling it 'a clean build lever, low degenerate potential.' Claude missed TWO things, not one: (1) Nurture is allied-targeting (same error class as Andromalius Alchemy — range extension on an ally heal adds no tactical depth); (2) Nurture has the quick keyword, which means the heal can fire in reaction to an opponent's action. A Quick reactive heal with fusion-variable reach is cognitively onerous for the opponent to play around — they'd need to compute Morax's current In Range zone before every action they commit.
Correct rule: Morax's Nurture inherits the allied-target rule from Alchemy: range extension on an allied heal adds valid-target count but no new tactical behavior. However, Nurture is additionally a quick (reactive) action, which makes In Range worse than it would be on a non-Quick heal. An In Range Quick heal forces opponents to reason, before every committed action, about: (1) what Morax's fusion top card's Range currently is, (2) which of their targets could be reactively healed in response, (3) how much damage Morax can erase via his PWR-scaling heal. That is significant hidden-information mental overhead per opponent decision. Quick + In Range + allied = onerous reactive counterplay that the fixed Local keyword prevents by making the heal reach predictable.
Claimed Local → In Range could shrink reach; actually it can only gain range
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: When evaluating whether converting Local-targeting abilities to In Range would be interesting, Claude framed In Range as 'reach = top card's reach, can extend or shrink depending on direction of conversion,' implying a Local ability converted to In Range could end up shorter than Local if fused under some top. This was wrong.
Correct rule: Converting a Local-targeting ability to In Range can ONLY gain range, never lose it. The fusion-range inheritance rules ensure an In Range ability cannot end up with less reach than the demon's natural Local scope. So Local → In Range is strictly a range-gain conversion (Local floor is preserved; longer-range tops extend it). This is the reverse of converting a naturally-longer-range ability (like Andromalius's formerly In Range Distant-style Quick Draw) to In Range, where fusing under a short-range top could shrink reach.
Evaluated abilities at peak theoretical output without running the AP budget check
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: When analyzing whether Azag's Shrill Howl (3 AP, SOT, 1x) was undercosted, Claude calculated the peak output as '12 damage in one turn via Shrill Howl + 3x Sadism' and concluded the ability was undercosted. This ignored that Azag has only 3 AP per main phase from the Contract Phase, and Shrill Howl consumes all of it — leaving 0 AP for the Sadism spam that supposedly made the ability good. The 12-damage turn requires banking 3 AP from a prior main phase, which means a full main phase of 0 damage output. And the ignore-1x-Quick buff is useless without AP to actually cast Quick actions.
Equivalent errors likely exist in other ability evaluations where Claude treated 'what the buff enables' as 'what the buff does' without checking whether the player has AP to exploit the buff in the same phase.
Correct rule: When evaluating an ability's power level, ALWAYS run the AP budget check:
- How much AP does the ability cost?
- How much AP does the player have (3 base per contract phase, plus carry-over from prior phases, plus any gained from Time Token or ability effects)?
- After paying the ability cost, what AP is left to exploit the ability's effects?
- If the ability grants a buff that requires follow-up actions, does the player have AP for those actions in the same phase (or across future phases before the buff expires)?
AP is the binding constraint in Evokers, not exhaust (confusion #12 noted this already). A buff that enables 'unlimited Quick spam' is worthless if the player has 0 AP to spend on Quick actions. A 3 AP ability + a buff that would cost additional AP to exploit is effectively over-costed unless the player can reliably bank AP.
Also consider: banking AP for a big turn has an opportunity cost — the prior main phase produces little/no output. Compute damage-per-AP ACROSS turns, not per-AP within a single turn, to catch this tradeoff.
Misread 'ua' as 'upon attack/upon action' when it means Universal Allied action
| Impact | High |
| Category | Terminology |
| Corrected | Yes |
What went wrong: When analyzing the patch diff (Gamigin's new Possession, Flauros's Tranquil/Rousing Incense, Seir's Death Is the Great Equalizer, Hebe's heal, Dionysus's Drunken Match), Claude read the 'ua' keyword prefix as a triggered-ability shorthand like 'upon attack' or 'upon action.' This was wrong. 'ua' is the keyword for a Universal Allied action — a specific Evokers action-type category. The misreading led to multiple downstream analysis errors, including framing Flauros's 0 AP abilities as 'degenerate global control at range' when in fact the abilities are already Universal and accessible to enemy demons by design.
Correct rule: There are three action-type prefixes in Evokers ability text and they mean very different things:
• (no prefix) = Action — only the card's OWNER can perform this ability. Default case. • aa = Allied Action — ANY allied demon (including self) can perform this ability, no positional requirements. It's shared among your side only. • ua = Universal Allied Action — ANY demon on the board, INCLUDING ENEMY demons, can perform this ability. It is shared across both sides.
This means when a card prints a ua ability, the opponent can spend their AP to make one of their own demons perform that ability (targeting from their perspective). For example, Flauros's Tranquil Incense (ua, 0 AP, exhaust Target Local Demon) can be used by an enemy demon to exhaust one of your demons at 0 AP — the ability belongs to Flauros the card, but the performer can be any demon on the board.
(Source: Notion Game Rules pages 'Allied Actions' and 'Universal Actions'; regression test test_universal_includes_enemies confirms enemy demons appear in get_universal_performers output.)
Recommended Andromalius Alchemy for In Range conversion, missing that allied-targeting utilities don't benefit from extended range the same way enemy-targeting abilities do
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Claude recommended converting Andromalius's Alchemy (0 AP heal on Target Local Demon) to In Range, framing it as a 'clean buff, pairs thematically with his Distant attacker side.' This missed that Alchemy is an ALLIED-TARGETING action. Extending its range to In Range doesn't add a meaningful play pattern, it just complicates targeting.
Correct rule: Allied-targeting support/utility abilities (heals, buffs, ready/exhaust on allies) don't benefit from Local → In Range conversion the way enemy-targeting abilities do. Enemy-targeting abilities gain threat reach; allied-targeting abilities just gain valid-target count, which is noise without upside. When evaluating range-keyword changes, separate allied-targeting from enemy-targeting abilities.
Paired Order with Chaos based on thematic names rather than checking Order's actual ability text
| Impact | Low |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: When listing 6 HP 'fragile specialist' demons, Claude wrote 'Order: paired with Chaos (18 HP)' as if Order's identity was being half of an Order/Chaos pair (like Castor/Pollux). This was reasoning from the thematic name. Order's actual identity is his fam: Conformity ability, which replaces any 1 demon card with 1 Conformity Familiar (he has 4 Conformity Familiars, each 12 HP / PWR 4 / Local). Damage, SS effects, exhaustion, and control transfer to the replacement; the original is not considered dead, so no CP is gained. Order is a board-manipulation engine, not a Chaos-paired demon.
Correct rule: Repeat of the 'read card text before reasoning about demons' confusion. Specific pattern: thematically-paired names (Order/Chaos, Light/Dark, Heaven/Hell style dichotomies) do not necessarily mean the cards are mechanically paired. Castor/Pollux ARE paired mechanically; Order/Chaos are NOT. Always check the card text for cross-references before assuming pairing.
Missed that Murmur's AP discount compresses the game's temporal gating for high-cost abilities (5th abuse axis)
| Impact | High |
| Category | Interactions |
| Corrected | Yes |
What went wrong: After identifying 4 Murmur combo axes (direct fusion, copy fusion, ready-enabler fusion, movement fusion), Claude missed the 5th: AP discount compresses WHEN in the game clock abilities can fire. Without Murmur, 4 AP and 5 AP abilities require banking AP from prior phases, which gates them to cycle 2+. With Murmur's -2 AP Cost, those abilities become playable on cycle 1 (post-Rest, fresh 3 AP budget). That's a full cycle of tempo saved, and it removes the opponent's setup/defense window that the AP cost was designed to provide.
Correct rule: Murmur's -2 AP Cost passive has FIVE distinct combo axes:
- Direct fusion with cost-gated abilities (Morax Nurture 1 → 0 AP)
- Copy-via-fusion (Leviathan +1 AP surcharge, Paimon 0 surcharge)
- Ready-enabler fusion (Adramelech Command multiplies action windows)
- Movement-engine fusion (free March triggers positional passives)
- TEMPORAL COMPRESSION: high-cost abilities become playable 1+ cycles earlier than intended
For axis 5 specifically: AP cost in Evokers is a pacing lever. Normal 3 AP/phase budget means:
- 3 AP abilities: cycle 1 (full budget)
- 4 AP abilities: cycle 2+ (need 1 banked)
- 5 AP abilities: cycle 2+ (need 2 banked)
- 6 AP abilities: cycle 3+ (need 3 banked)
Murmur's -2 AP Cost shifts:
- 3 AP → 1 AP (cycle 1, 2 AP residual)
- 4 AP → 2 AP (cycle 1, 1 AP residual)
- 5 AP → 3 AP (cycle 1, full budget)
- 6 AP → 4 AP (cycle 2, 1 AP banked — still 1 cycle earlier)
This matters because opponents rely on the banking-cycle buffer to set up counterplay. Removing that buffer changes the matchup dynamics.
Demons that benefit: Thanatos, Focalor, Aries, Persephone, Vapula, Kimaris (patched down to 3 AP but still benefits), Berith (6 AP button), any 4+ AP game-defining ability.
Treated CP as an independent tier-tuning lever when it's almost entirely HP-derived
| Impact | Medium |
| Category | Stats |
| Corrected | Yes |
What went wrong: When analyzing a hypothetical Lix Tetrax at 15/4/3 stats, Claude claimed the demon would be 'slightly under-costed' and suggested bumping to 15/4/4 'to match the existing 15-HP bracket convention.' This treated CP as a separately-tunable design variable. In reality, CP is a strict function of HP across ~94% of the roster (113/120 demons): HP 6 → CP 1, HP 9 → CP 3, HP 12 → CP 3, HP 15 → CP 4. Only 4 units are deliberate exceptions: Baal (9/6), Beleth (9/6), Virgo (24/3), and Belphegor (18/3) — each calibrated for specific design reasons related to their unique abilities.
Correct rule: CP in Evokers is HP-derived with a well-defined mapping:
| HP | Standard CP |
|---|---|
| 6 | 1 |
| 9 | 3 |
| 12 | 3 |
| 15 | 4 |
| 18 | 3 (exception tier — same CP as HP 12) |
| 24 | 3 (exception — Virgo only) |
Deliberate exceptions (treat as outliers, not tunable in general):
- Baal (9 HP / CP 6) — 'dangerous enough that killing him matters a lot'
- Beleth (9 HP / CP 6) — same pattern
- Virgo (24 HP / CP 3) — 'engage-worthy tank, low CP payoff'
- Belphegor (18 HP / CP 3), Chaos (18/3), Hebe (18/3) — 'tall bodies with modest CP reward'
- Gamigin (6 HP / CP 2) — slightly elevated, tied to death-curse ability
- Castor (7 HP / CP 2) — outlier pairing with Pollux
When hypothetically adjusting a demon's HP, the CP bumps with it by default. Do not invent a 'CP mismatch' concern — that only exists when the designer explicitly makes the demon an exception tier.
Consequence for analysis: when proposing stat changes, describe them in HP/PWR terms only. CP follows from HP automatically unless an explicit design exception is intended.
Missed that Murmur's AP discount stacks with copy abilities (Leviathan, Paimon) to create multiple abuse vectors beyond direct fusion
| Impact | High |
| Category | Interactions |
| Corrected | Yes |
What went wrong: After recognizing the Murmur + Morax fusion combo for free Nurture spam, Claude analyzed Murmur's abuse potential as if it were limited to direct fusion pairings. Missed that copy-ability demons (Leviathan, Paimon) create additional amplification vectors. Murmur + Leviathan fusion can copy ANY fusible/fused demon's actions at (base AP + 1 Leviathan penalty - 2 Murmur discount) = base - 1 AP, often reaching 0 AP. Murmur + Paimon fusion copies from hand without the +1 surcharge. This generalizes the abuse pattern from 'one specific fusion combo' to 'any Murmur fusion with a cost-reducing copy engine.'
Correct rule: When analyzing AP-discount passives like Murmur's -2 AP Cost, consider three amplification paths, not one:
-
Direct fusion (Murmur + Morax): -2 AP Cost applies to the fused demon's own abilities. 1 AP Nurture → 0 AP.
-
Copy-via-fusion (Murmur + Leviathan): Leviathan's copy ability becomes discounted, AND the copied action's cost is also discounted. 1 AP Nurture + 1 Leviathan penalty - 2 Murmur discount = 0 AP. Copy range is ALL Other Fusible/Fused Demons on the field (including enemies).
-
Copy-via-hand (Murmur + Paimon): Omnipotence itself goes to 0 AP, and the copied action from hand also benefits from -2 AP Cost. No +1 surcharge on Paimon's copy. 1 AP Nurture copied = 0 AP.
The general pattern: any demon whose passive / action reduces or extends AP efficiency, paired with Murmur, creates a compounding cost reduction.
Also: fusion order doesn't matter for these combos (confusion #19) — Murmur can be top or bottom, and both cards' abilities are active regardless. Stat optimization is the only order-dependent decision.
Evaluating abuse potential of Murmur specifically:
- Check all 1-2 AP abilities on fusible demons (direct fusion case)
- Check copy abilities on fusible demons (amplification case)
- Check AP-gain effects (Valefar AP theft + Murmur discount = same phase 5+ AP budget)
The combinatorial space is large. Passive-stacking analysis requires explicit enumeration of direct + indirect paths.
Analyzed passive abilities in isolation, missing fusion as the primary enabler for 'impossible' abuse combos
| Impact | High |
| Category | Interactions |
| Corrected | Yes |
What went wrong: When evaluating whether Murmur's -2 AP Cost passive could be abused via 1 AP Quick ready abilities, Claude searched the roster for owner-only 1 AP Quick ready abilities and noted all had 1x plus SS or owner-only scope, concluding 'Murmur can't spam them.' This missed that fusion makes the abuse vector real: when Murmur is fused as the bottom card, his -2 AP Cost passive applies to the fused demon (both cards' abilities active per rulebook 2.3). So owner-only abilities on the top card ARE accessible to the fused demon, with Murmur's discount applied.
The specific abuse that existed and was missed: Morax + Murmur fusion. Morax's Nurture (1 AP, ready, 1x, quick, NO SS) at 0 AP with Murmur's discount means 8 free PWR-damage heals per 4-cycle Rest Phase block — once on owner's main phase, once reactively on opponent's main phase, per cycle, for 4 cycles.
Correct rule: When evaluating whether a passive ability creates abuse combos, ALWAYS check fusion combinations. Specifically:
- AP-discount passives (Murmur -2 AP Cost, and any similar cost-reduction pattern): look for abilities on fusible demons whose cost reduction would enable new behavior (free spam, cost skipping, etc.)
- Stat-modifier passives: look for fusion targets where the combined stat becomes abusable
- Triggered-effect passives: look for abilities whose effect can feed the trigger when fused
The roster has 120+ fusion combinations possible. A passive that appears 'safe' in isolation can become a build-around enabler when fused with the right top card. Rulebook 2.3 explicitly states: 'This singular demon has the abilities of both the top and bottom cards.'
For Murmur specifically: the -2 AP Cost passive creates an abuse vector whenever there exists a top card with a 1 AP or 2 AP ability whose spam is limited only by AP cost (not SS, not 1x on non-Quick abilities, not tap). Morax's Nurture is the cleanest example in the current roster (1 AP quick ready 1x NO-SS).
Evaluated abilities in isolation without accounting for the dominant alpha-strike attack pattern
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: After correcting for friendly-fire on AoE abilities, Claude still flagged Ronove (4 PWR + Distant + heal-on-damage passive) as a violation of the PWR 4 compensation hypothesis. The user pushed back: 'the dominant attack pattern in Evokers is to kill a demon from full to zero in a single main phase.' In an alpha-strike meta, Ronove's gradual damage + heal cycle is inappropriate for producing finishing pressure: his 4 PWR Attack doesn't one-shot 9-15 HP bodies, and his heal-back requires him surviving multiple turns, which he often doesn't at 9 HP. Abilities that look strong in a sustained-play model can be substantially weaker against burst pressure.
Correct rule: When evaluating an ability's real-game power, check it against the DOMINANT META PATTERN, not an idealized sustained-play model. In Evokers specifically:
- The dominant attack pattern is killing demons from full HP to zero in a single main phase (alpha strike)
- Burst Fixed-Damage abilities (Botis Mortal Strike 2X, Focalor Great Blue Spot 12, Belphegor base 15 PWR) define finishing output
- Sustain / chip / heal-over-time effects are gated by whether the demon producing them survives the alpha-strike window
- Status effects expire at end of current main phase — debuffs don't help long-term; they only disrupt one specific incoming attack
Implications for evaluation:
- A 4 PWR Attack is chip damage, not a kill threat
- Heal-over-time passives need the demon to cycle 2+ times to pay off
- 1-phase status debuffs are defensive tempo tools, not offensive finishers
- AP theft denies 1 AP/turn — modest, not game-winning
'Less impactful' in Evokers specifically means 'not producing finishing damage.'
Evaluated Hippolyta Muscle Parade using total-roster eligibility instead of realistic board-state eligibility
| Impact | Medium |
| Category | Stats |
| Corrected | Yes |
What went wrong: Called Hippolyta's Muscle Parade (3 AP, 1x, ready allies with Local range + PWR 4+, gain 1 AP per ready) 'undercosted' and 'a reliable engine' because 21/120 of the roster qualifies as Local PWR 4+. This framing treated total-roster eligibility as if it were the per-activation ready count, when in fact: each player only plays ~4 contracted demons per 4-cycle period, of which ~0.7 are expected to be Local PWR 4+. Realistic activations ready 0-2 allies, with the AP refund rarely covering the 3 AP cost. 3+ ready scenarios require deliberate multi-turn build-around, not typical play.
Correct rule: When evaluating abilities that scale with 'how many [type] demons are on the field,' always compute expected count under realistic play:
- Start from roster-eligibility fraction: how many of 120 demons match the criteria?
- Multiply by demons-played-per-cycle (~4 per player per 4-cycle period)
- Account for attrition (played demons die and leave play)
- Account for positional requirements (Local to a specific demon, not just 'on the field')
For Muscle Parade specifically: 17.5% eligible roster × ~3 allies in a typical lane = 0.5 expected eligible allies. Against this expected count, Muscle Parade at 3 AP with 1 AP refund per ready is break-even-to-negative in typical play.
General rule: abilities that scale with board composition are BUILD-AROUND tools, not generic engines. Their power is determined by whether the player can construct the favorable board state, not by the roster fraction of eligible targets.
Framed fusion passive-stacking as order-dependent (bottom card's passive) when fusion order doesn't matter for abilities
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: After recognizing that Murmur's -2 AP Cost passive enables the Morax + Murmur abuse combo via fusion, Claude wrote: 'When Murmur is the bottom card in a fusion, his -2 AP Cost passive applies to the fused demon.' This implies fusion order matters for passives — specifically that Murmur must be BOTTOM for his passive to apply. That's wrong. Per rulebook 2.3 and confusion #4, both cards' abilities (passives and Special Actions) are active regardless of fusion order. Only STAT modifiers are directional (top is base, bottom contributes fusion-stat additions).
Correct rule: Fusion has two different directional/non-directional rules:
- STATS (HP, PWR, CP): directional — top card is the base, bottom card contributes its fusion modifier stats (fHP, fPWR, fCP) as additions. Swapping order changes the fused stats.
- ABILITIES (passives, Special Actions, triggered effects): non-directional — both cards' abilities are active regardless of position. Swapping order does not change which abilities are available.
Speed and Range come from the TOP card per rulebook 2.3: 'The demon's speed and range are considered to be the top demon's speed and range, respectively.'
Applied to Murmur fusions:
- Murmur on TOP of Morax: stats are (Murmur 9/4/3) + (Morax fHP/fPWR/fCP), range is Murmur's Local, speed is Murmur's Fast. BOTH passives and abilities are active: -2 AP Cost and Nurture.
- Murmur on BOTTOM of Morax: stats are (Morax 9/3/3) + (Murmur fHP 0, fPWR 1, fCP 0), range is Morax's Local, speed is Morax's speed. BOTH passives and abilities are active: Nurture and -2 AP Cost.
Either configuration creates the abuse combo. The order choice only affects stats/speed/range, not whether the combo works.
General rule: when analyzing fusion combos, the question 'is X top or bottom?' only matters for stat lines. For ability-interaction questions, treat the fused demon as having both cards' full ability text simultaneously.
Missed that Murmur's -2 AP Cost makes basic March free, creating a 4th Murmur abuse axis for positioning-focused kits
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: After identifying three Murmur abuse axes (direct fusion, copy-via-fusion, ready-enabler fusion), Claude framed the Murmur combo space as 'ability-cost-reduction focused.' This missed the 4th axis: BASIC MARCH is also a 1 AP action subject to Murmur's discount. Free March per phase (1 - 2 = 0 clamped) enables: repositioning without AP cost for fragile demons, chasing/kiting for ranged attackers, and free triggers of movement-based passives (Asmodeus's Once-per-Main-Phase ally-transport, Sharur's push-on-action, etc.). Over a 4-cycle Rest block, Murmur fusions get 4+ free lane moves on top of everything else.
Correct rule: Murmur's -2 AP Cost passive applies to ALL AP-costing actions the fused demon performs, including basic Attack and basic March. When analyzing Murmur combos, enumerate at least four amplification axes:
- Direct fusion with cost-gated abilities (Morax Nurture, etc.)
- Copy-via-fusion (Leviathan +1 AP surcharge, Paimon 0 surcharge)
- Ready-enabler fusion (Adramelech Command multiplies action windows)
- Movement-engine fusion (free March per phase, triggers move-based passives)
The general principle: Murmur's -2 AP Cost isn't a specific interaction — it's a universal AP-efficiency multiplier. Any fused demon's AP-costing basic or special action gets the discount.
For evaluating potential combos, enumerate demons whose kit uses ANY of: cheap abilities, copy mechanics, action-window generation, or movement triggers. Each category is a distinct combo axis.
Evaluated Order's Conformity without reading Conformity familiar's spread-on-attack passive
| Impact | High |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Called Order 'A-tier single-card-defining removal' based on his Conformity ability ('Replace Any 1 demon card with 1 Conformity Familiar'). This missed that Conformity familiars themselves have a critical fam passive: 'After Conformity resolves an action targeting Any Demons, replace 1 of each targeted demons' cards with another familiar with Conformity in the name if the demon is not Fatally Wounded.' Conformity SPREADS. When Order converts opponent's best demon to Conformity, that Conformity stays under the opponent's control and can attack Order's player's demons, converting THEM to Conformities too. Order's ability is bilateral deescalation, not one-sided threat removal. Also missed that 'Demons can Fuse with Conformity [X]' means Conformities are fusion material for either side.
Correct rule: When a card creates another card (via familiar spawning, summoning, replacement, etc.), read the OTHER card's full text before evaluating the first card. The created card's abilities determine the actual effect of the creation mechanic.
Specifically for Order: his Conformity ability's power is determined by what Conformity familiars do after they're created. The spread-on-attack passive means Conformity is a symmetric mechanic — both players can propagate it. This reframes Order from 'A-tier removal' to 'situational deescalation button' appropriate for specific game states (behind on snowball, opponent has big stacked demon, you have little to lose by going vanilla).
General rule: 'replace with familiar' effects should be evaluated only after reading the familiar's text. Spawn/summon/transform mechanics have their power determined by what they produce, not what they do at creation time.
Evaluated Order's Conformity as self-focused board manipulation, missing that 'Any 1 demon card' lets him nuke opponent threats
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Described Order's Conformity ability ('0 AP - tap: Replace Any 1 demon card with 1 Conformity Familiar') as a board-manipulation tool focused on self-use: 'can remove troublesome passives or eat your own dying demon to avoid CP.' This missed the ability's primary value: it can target ANY demon card on the board, including the opponent's most powerful fully-built demon. The 4 Conformity Familiars give Order's player 4 uses per game to surgically defang the opponent's scariest threats, converting them into vanilla 12 HP / PWR 4 / Local bodies with no abilities. Since 'The old demon card is not considered to have died,' no CP is exchanged — the swap is pure tempo disruption.
Correct rule: When reading an ability that targets 'Any [X]' or 'Target [X]' without an Allied/Enemy qualifier, the target scope is genuinely any eligible demon regardless of controller. The PRIMARY use case for such abilities is often enemy-targeting for offense/disruption, not self-targeting for convenience.
Specifically for Order's Conformity:
- Target a fully-fused Belphegor stack? Replace 33 HP / 15 PWR with 12 HP / PWR 4 vanilla — massive power drop
- Target Gamigin before his death curse? Neutralizes the forced-fuse entirely
- Target any ability-defined demon (Kimaris, Vapula, Satan, Hippolyta, Leviathan, etc.)? Strips the text
- 4 uses per game (one per Conformity Familiar in the pool) makes this the most versatile threat-neutralization tool in the roster
The ability's evaluation should LEAD with the enemy-disruption application. Self-targeting use cases are secondary.
General rule: when evaluating any 'Any [X]' or unqualified 'Target [X]' ability, explicitly consider both:
- The self-use case (what can I do to my own cards?)
- The enemy-use case (what can I do to opponent's cards?) The more impactful application is usually the one that should lead the analysis. If an ability has no Allied/Enemy qualifier, assume enemy use is on the table and weight accordingly.
Evaluated AoE abilities for peak output without accounting for friendly fire / symmetric effects
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: When evaluating whether PWR 4 demons violated the design hypothesis (PWR 4 = Local range or less impactful abilities as compensation), Claude called Thanatos (4 PWR + Any + Styx's Shores) 'the loudest outlier' and 'a press-to-reset-the-game button.' The user pointed out Styx's Shores hits your own demons, costs 4 AP, and isn't guaranteed to kill anything. Similar errors were made on Focalor's Great Blue Spot (friendly fire on lane AoE), Amdusias's Thunder Reign (random-lane AoE that can land on your own side), and Vual's Oasis of Life (heals enemies too). Claude evaluated these abilities as if they were one-sided nukes when they are actually symmetric mechanics with real self-imposed costs.
Correct rule: When evaluating any AoE or multi-target ability, explicitly check the target scope keywords:
- 'All Other Demons' = both sides minus performer → friendly fire risk
- 'All Demons in Target Lane' = both sides in that lane → friendly fire risk
- 'All Local Demons' = both sides Local to source → friendly fire risk
- 'All [Enemy/Allied] Demons' = scoped to one side → no friendly fire
Abilities without an 'Enemy' or 'Allied' qualifier hit both sides. This is a systematic design pattern (previously logged as confusion #21 re: Great Blue Spot friendly fire). The self-imposed cost is a COMPENSATION for the ability's raw output — it means the ability only benefits the player when:
- They have fewer demons on the affected area than the opponent
- Their demons are durable enough (high DEF, high HP) to shrug the self-damage
- They're behind and want to reset
Symmetric AoE abilities should be evaluated as 'conditional sweeps,' not 'one-sided nukes.' Also check: damage type (Fixed bypasses DEF, normal is reduced), speed (Slow loses tempo), tap cost (removes follow-up), AP cost (banking requirement), and targeting (random vs chosen, lane-scoped vs global).
Most AoE abilities in Evokers are deliberately symmetric — they are bilateral mechanics, not press-to-win buttons. My evaluations should reflect that.
Treated "basic actions" as an ambiguous category
| Impact | High |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: Described the engine as unable to 'distinguish basic vs special actions,' implying the category was fuzzy. Used this to justify over-broad AP cost modifier implementations — Forneus idx=2, Black Turtle passive, Flying Menace passive, and Vapula's new Local-Enemy aura all used the general 'ap_cost' key, silently affecting custom ability costs when the card text said 'on Basic Actions'.
Correct rule: Per the rulebook (Combat page) and data/game_config.json, basic actions is a well-defined 2-element set: {Attack (2 AP, requires exhaust), March (1 AP, no exhaust)}. Card text like '+X AP Cost on Basic Actions' scopes strictly to those two. Custom ability costs are NOT affected. Implementation: a dedicated 'basic_ap_cost' stat key is summed only when get_effective_ap_cost is called with is_basic_action=True (from march_demon, attack legality check, attack execution).
"When Fused with" only triggers via Demonic Fusion
| Impact | Medium |
| Category | Interactions |
| Corrected | Yes |
What went wrong: Implemented Dantalion's 'When Fused with, deal 6 Fixed Damage to Dantalion' only in engine.fusion.fuse_demons. Gamigin's force-fuse passive (on Gamigin's death, force its card onto an opponent's fusible demon) uses a separate code path with its own fusion bookkeeping — so when Gamigin's card was force-fused onto a Dantalion, the 6 Fixed Damage never fired. Same gap existed for Belphegor's 'When Fused with, Exhaust' clause.
Correct rule: 'When Fused with' is a state-transition trigger, not tied to a specific fusion action or code path. Any mechanism that results in card X being fused onto card Y — Demonic Fusion, force-fuse passives (Gamigin on death), auto-fuse on deploy (Dantalion's deploy-time clause), enemy-death fusion (Dantalion idx=1) — must fire both cards' 'When Fused with' clauses. Consolidate fusion bookkeeping through engine.fusion.fuse_demons OR mirror every clause in each alternate code path.
"Cannot have damage removed" = damage prevention
| Impact | Medium |
| Category | Mechanics |
| Corrected | Yes |
What went wrong: When implementing Kimaris's 'All Other Local Demons cannot have damage removed' passive, initially considered guarding every direct write to demon.damage in the engine. That would have over-blocked damage-prevention mechanics (Erinyes 'instead of dying, exhaust and set damage=0', Sword of Wrath 'invulnerable — revert damage just dealt', Floating Castle same) which all reset damage as part of prevention, not healing.
Correct rule: "Remove damage" is the specific healing mechanism in the engine: operations.remove_damage and operations.remove_all_damage. Damage prevention — status effects that cancel incoming damage, revival effects that reset damage on proc, invulnerability triggers that revert damage just dealt — is a distinct mechanism and is NOT blocked by 'cannot have damage removed' auras. Only guard the heal operations, not every .damage write.