Great League PvP-Ranking

The PvP factory does not reproduce a community or usage ranking. It creates its own simulation ranking from the local GameMaster data, calculated league IVs, and precomputed matchup simulations.

First, the candidate pool is built per league. A Pokémon is included only if it has values legal for that league: Great League uses the stored Great League IV, Ultra League uses the stored Ultra League IV, and Master League uses a 15/15/15 level-50 build. Unreleased Pokémon, temporary forms, blacklist entries such as Mew, and Pokémon below the minimum CP filter are excluded.

Then the factory creates every technically legal moveset build for each Pokémon: every legal Fast Move combined with every pair of two legal Charged Moves. Elite moves are included when the run enables Elite moves. Only technically incompatible moves, duplicates, and globally excluded PvP moves are removed.

Every build is simulated as its own unit. The main categories are Lead, Safe Switch, and Closer. They differ by shield and energy state. The simulator plays out the 1v1 situation and stores the result score for both sides.

The matchup score is HP-based: own remaining HP percent plus 100 minus the opponent's remaining HP percent. A clear win with 60 percent remaining HP scores 160 points; a narrow loss against an opponent with 5 percent remaining HP scores 95.

After simulation, the recursive ranker scores builds first, not Pokémon directly. A build is evaluated against all opposing species. Within an opposing species, builds are aggregated with weights so species with many legal movesets are not over-weighted.

Opponent weighting is recursive. In each iteration, stronger opposing Pokémon receive more weight and weaker ones receive less. This makes a good result against a strong meta Pokémon count more than a good result against a weak Pokémon.

Within a Pokémon, multiple builds are combined via softmax. The strongest build counts most, while weaker builds count less without being fully discarded. The visible placement follows this recursively opponent-weighted score, not pure win rate, usage frequency, or an external community list.