The Pokémon Company has announced a new experience called "AI Battle Challenge," which is now pitting human players against its bots in the "Pokémon Trading Card Game."
The goal is to see how humans fare against existing AI bots, but more importantly, train their existing models on how humans play the game, their strategies, and more.
'Pokémon TCG' 'AI Battle Challenge' Is Here
The "Pokémon TCG" "AI Battle Challenge" is now live, and it is asking programmers from around the world to help solve a problem that has frustrated casual and competitive players alike for years.
The "AI Battle Challenge" officially kicked off on June 16 and will run through August 17, offering two unique challenge categories for participants. The goal of the competition is for teams to train a battle simulator AI agent and test it against other submissions in a tiered daily ladder system.
Final entries are due August 16, with The Pokémon Company planning to run games through the end of August to determine the final leaderboard.
Anyone interested in participating can find the full rules and submission guidelines on the official "AI Battle Challenge" website. The winning team will be whoever produces the agent that finishes in first place on the final leaderboard.
Humans vs. Bots in 'Pokémon TCG' Competition
According to ComicBook Gaming, the competition structure pits AI agents against each other rather than directly against human players during the evaluation phase.
However, the data and techniques gathered from the competition are intended to raise the bar on what the bots can do in actual game modes. The core challenge that the competition is trying to address is a well-documented one.
Anyone who has played solo battles in "Pokémon TCG Live" or "Pokémon TCG Pocket" has likely noticed that the computer opponents make puzzling decisions that no experienced player would make in a real match.
The reason AI has such a hard time with "Pokémon" card games comes down to hidden information. In chess, a game where AI has famously dominated human players for decades, both players can see the entire board at all times.
In a "Pokémon" card battle, neither player can see the other's hand, which forces a more adaptive and probabilistic style of play. That makes it significantly harder for a machine learning model to anticipate and respond to what an opponent might do next.
The Pokémon Company Wants to Train AI Models
The "AI Battle Challenge" aims to enhance the performance of an AI training agent within the "Pokémon Trading Card Game" by having competitors build and submit their own models for evaluation. The framing of the competition makes clear that this is about developing better AI-driven opponents, not integrating generative AI into the card game itself.
If The Pokémon Company uses the results to improve its in-game AI battles, players in "Pokémon TCG Live" and "Pokémon TCG Pocket" could eventually face opponents that are meaningfully harder to beat in solo and training modes.
This feature could be a welcome improvement for newer players using solo battle modes to sharpen their skills before taking on human opponents online.
