Pokémon Showdown is an open-source simulator that transforms Pokémon's turn-based battles into a competitive strategy game enjoyed by thousands of daily players. Competitive Pokémon battles are two-player stochastic games with imperfect information. Players construct teams of Pokémon and guide them through complex battles where stochastic outcomes are driven by nuanced gameplay mechanics. Success requires more than managing randomness; players must effectively model their opponents and make decisions amid uncertainty. Essential details of the opponent’s team remain concealed until they impact the battle, encouraging players to infer hidden information and predict future moves based on past interactions. Expert human players distinguish themselves by accurately anticipating their opponent's strategies. The stochasticity, partial observability, and diverse team options of Pokémon battles challenge AI's ability to plan and generalize.

Though Pokémon Showdown battle bots have existed for many years, advances in language models, large-scale reinforcement learning datasets, and accessible open-source tools have sparked renewed interest within the machine learning research community. Recent methods have achieved human-level gameplay in popular singles rulesets, prompting an exciting question: How much further can we push the capabilities of Competitive Pokémon AI? Join Track 1 of the PokéAgent Challenge and help us find out!