Series

Pokémon

3 games, 55 claims. Each cell lists the claims about that concept in that game; “uses” marks a game that has the concept but no claims yet.

Concept	Pokémon Red and Blue 1996	Pokémon Crystal 2000	Pokémon Emerald 2004
Damage formula	Type picks the stats: Fire, Water, Grass, Electric, Psychic, Ice, and Dragon (`SPECIAL` upward) use Special on both sides; the rest use Attack against Defense. T3 In Red and Blue, a damaging move runs `CriticalHitTest`, `CalculateDamage` (base formula), `AdjustDamageForMoveType` (STAB, then type rows), `RandomizeDamage`, and then `MoveHitTest` (accuracy). T3 `CalculateDamage` computes $⌊ ⌊ (⌊ 2 \times Level / 5 ⌋ + 2) \times Power \times Attack / Defense ⌋ / 50 ⌋$ , caps it at 997, and adds 2, so the type step receives 2–999. T3 `RandomizeDamage` multiplies damage by a random byte, 217–255, and integer-divides by 255, leaving 0 or 1 damage untouched. T3	Type still picks the stats: Steel is physical (below `SPECIAL`); Dark is special, using Special Attack against Special Defense. T3 A plain damaging move runs critical, damagestats, damagecalc (base formula, item boosts, and critical ×2), stab (weather, badges, STAB, and type rows), damagevariation, and then checkhit (accuracy). T3 The random factor is unchanged from Red/Blue, a random byte from 217 to 255 over 255. T3	Type still decides physical or special in Emerald, by its position relative to `TYPE_MYSTERY`. T3 A plain damaging move runs critcalc, damagecalc (base damage), typecalc (STAB, then type rows), and then adjustnormaldamage (random factor). T3 The random factor changes form: damage × (100 − rand % 16) / 100, i.e. 85–100 %, raised to 1 if that gives 0. T3
Same-type attack bonus	`AdjustDamageForMoveType` applies STAB before any type row: $damage + ⌊ damage / 2 ⌋$ when the move's type matches either attacker type. T3	`BattleCommand_Stab` applies weather and badge type boosts, then STAB ( $damage + ⌊ damage / 2 ⌋$ if either attacker type matches), then the type chart. T3 A comment in pokecrystal's `BattleCommand_Stab` expands STAB as “Same Type Attack Bonus”. T3	Before the type chart, `Cmd_typecalc` applies STAB as $damage \times 15 / 10$ when either attacker type equals the move's type. T3
Type effectiveness	Red and Blue's type chart (`TypeEffects`) has 82 non-neutral (attacker, defender, multiplier) rows over 15 types, ended by -1; unlisted pairs are neutral. T3 Multipliers are integers scaled by 10 (`SUPER_EFFECTIVE` 20, `NOT_VERY_EFFECTIVE` 05, and `NO_EFFECT` 00); the table never uses `EFFECTIVE` (10) or `MORE_EFFECTIVE` (15). T3 In table order, each `TypeEffects` row matching the move's type and either defender type sets damage to $⌊ damage \times value / 10 ⌋$ ; 0 marks a miss. T3 Under two ½ rows (×¼), 2 or 3 damage truncates to 0 and the move misses; 4 damage gives 1. T4 Truncating after each row costs a net-neutral hit damage when its ½ row comes first: Ice against Water/Flying takes 3 to ⌊3 × 5 / 10⌋ = 1, then ⌊1 × 20 / 10⌋ = 2. T4 The Red/Blue type chart gives Ghost-type moves no effect on Psychic types. T3 In Red and Blue, a Saffron Gym trainer's after-battle line contradicts the table: “Psychic #MON fear only ghosts and bugs!” T3 Bulbapedia reports that the Gen 1 Ghost → Psychic ×0 “may have been an error”, citing Nintendo strategy guides, two anime episodes, and the Saffron Gym line. T1 Each matching row overwrites the byte `wDamageMultipliers`, so the text follows the last row: above 10 is “super effective”, below 10 “not very effective”, and a miss reads “doesn't affect” only at 0. T3 Since the text follows the last row, an Ice move against Water/Flying, net ×1 (½ row at L33, then 2× at L36), prints “It's super effective!”. T4 A Ground move against Rock/Flying (×0 row at L9, then ×2 at L55) does 0 damage and prints the plain miss text, not “doesn't affect”: the byte holds the last row's 20. T4 `SetDamageEffects` moves (Sonic Boom, Seismic Toss, Dragon Rage, Night Shade, Psywave, and Super Fang) jump to the hit test, skipping `CalculateDamage` and `AdjustDamageForMoveType`, the only damage routine that reads `TypeEffects`. T3 The trainer AI's `AIGetTypeEffectiveness` stops at the first row matching either defender type, ignoring dual-type products. T3 The type constants still define `BIRD` (`$06`), with a name string, but no `TypeEffects` row names it. T3	Crystal's type chart (`TypeMatchups`) has 108 non-neutral rows over 17 types, then a -2 marker, Normal→Ghost and Fighting→Ghost ×0, and a -1 end. T3 Multipliers are unchanged from Red/Blue: scaled by 10, with `SUPER_EFFECTIVE` 20, `NOT_VERY_EFFECTIVE` 05, `NO_EFFECT` 00, and `EFFECTIVE` 10. T3 Crystal's type constants add `STEEL` (physical), `DARK` (special), and an unused `CURSE_TYPE` named “???”; `BIRD` remains, still with no rows. T3 Struggle returns from `BattleCommand_Stab` at once, so it skips weather, badge boosts, STAB, and the type chart. T3 Each matching row computes $damage \times value$ ; a nonzero product becomes $⌊ product / 10 ⌋$ , raised to 1 if that truncates to 0. T3 A ×0 row now marks the attack as missed outright, instead of relying on the damage reaching 0. T3 A ×¼ hit no longer truncates to a miss: 2 damage through two ½ rows gives ⌊2 × 5 / 10⌋ = 1, then ⌊1 × 5 / 10⌋ = 0, raised to 1. T4 Against an identified (Foresight) target, the loop stops at the -2 marker, skipping the Normal→Ghost and Fighting→Ghost immunities after it. T3 After the loop, `CheckTypeMatchup`'s combined value ( $m = ⌊ m \times row / 10 ⌋$ from 10) replaces the effectiveness byte, so the text reflects every row: ×4 is 40, ×¼ is 2, and ½ with 2× is 10. T3 An Ice move against Water/Flying (½ at L35, 2× at L39) prints no effectiveness text: the combined value goes 10 → 5 → 10. T4 Row order still matters: an Ice move against Water/Flying takes 3 damage to 1, then 2 (½ row at L35, 2× at L39). T4 Crystal's type chart makes Ghost-type moves super effective on Psychic types. T3 Crystal's table adds Steel and Dark and changes four Red/Blue matchups: Poison→Bug ×2 → ×1, Bug→Poison ×2 → ×½, Ghost→Psychic ×0 → ×2, and Ice→Fire ×1 → ×½. T4	Emerald's type chart, `gTypeEffectiveness`, is a flat u8 array of 108 non-neutral (attacker, defender, multiplier) rows over 17 types, then a `TYPE_FORESIGHT` marker, two Ghost immunities, and `TYPE_ENDTABLE`. T3 Emerald keeps the Game Boy encoding, now named `TYPE_MUL_*`: 20 super effective, 10 normal, 5 not effective, and 0 no effect. T3 Emerald's table holds the same 110 rows as Crystal's, in the same order: no new type and no changed matchup. T4 `TYPE_MYSTERY` (9), named "???", sits between the physical and special types and has no rows. T3 For each row of the move's type, `ModulateDmgByType` runs once per distinct target type it matches: $damage \times multiplier / 10$ , raised from 0 to 1 unless the multiplier is 0. T3 The text comes from flags `ModulateDmgByType` sets per row: a 2× row cancels an earlier not-very-effective flag, a ½ row an earlier super-effective one, and ×0 sets "doesn't affect", clearing both. T3 Against a Foresight target, the loop breaks at the `TYPE_FORESIGHT` row, skipping the Normal→Ghost and Fighting→Ghost immunities after it. T3 Struggle skips `Cmd_typecalc` entirely, so neither STAB nor the type chart touches it. T3 `Cmd_typecalc` flags a Ground move against a Levitate target as missed and “doesn't affect”, and walks no type row. T3 After the rows, a Wonder Guard target makes a damaging move miss on the turn it attacks, unless the move carries the super-effective flag without the not-very-effective one. T3 Row order still matters: an Ice move against Water/Flying takes 3 damage to 1, then 2 (½ row at L369, 2× at L373). T4