Raid counter ranking methodology

Pure DPS rankings (like PokeBattler's simulator output) miss two important factors: how often the attacker is meta-relevant across many bosses, and how reliably it survives the boss's charged moves.

A glass cannon with 350 DPS that faints in 6 seconds is worse in practice than a balanced attacker with 280 DPS that survives 18 seconds. Both matter, and the ranking combines them.

score = tier_weight × STAB × type_effectiveness × moveset_dps × bulk

tier_weight    = lookup(tier)            // S=100, A+=80, A=60, B=40, C=20
STAB           = 1.2 if move type ∈ attacker types else 1.0
type_eff       = product over boss types  // 1.6 SE, 0.625 res, 0.39 imm
moveset_dps    = published top-moveset DPS for the attacker
bulk           = (HP × Def)^0.5 / 10000   // normalized survival proxy

PokeBattler runs a full Markov-chain simulation of the raid fight, which is more accurate but takes ~100ms per attacker × moveset. We can't do that for 1 200 mons × 5 movesets each at request time on a static site.

Instead we use a closed-form score that captures 90% of the ranking accuracy in 1ms per attacker. PokeBattler is the ground truth — we benchmark against it quarterly and adjust weights if the rankings drift.

Each boss page shows the top 12 attackers by score. We pick 12 because:

• ▸You bring 6 attackers per raid (3 mons × 2 trainer slots in a duo).
• ▸Top 12 = enough variety to handle most movesets the boss can throw at you.
• ▸Beyond 12, score drop-off is steep — #13 is usually <80% of the #1 score.