AI chess on ChessMoves AI

ChessMoves AI puts three kinds of AI to work for one purpose: helping you play better chess. A chess engine finds the moves, a model of human play predicts what a person of a given rating would play, and a language model explains what happened in plain words, checked against the engine.

Three kinds of AI, each doing what it is good at

A chess engine (Stockfish) evaluates positions and finds the best moves. It decides which of your moves were mistakes and by how much.
A model of human play (Maia-2), trained on millions of human games, predicts the move a player of a given rating is most likely to make. It powers move prediction and the simulated opponents in opponent preparation.
A language model turns the engine's findings into explanations a person can use. The moves it names are checked against the board, so an explanation never recommends a move the engine did not find.

Analyse your games

Pick a game you played on ChessMoves AI, or upload a PGN of games played anywhere else, and every game gets a move-by-move review: the mistakes, how much each one cost, what you should have played, and why. See AI chess analysis.

Get coached on your real weaknesses

Across your games, the AI chess coach finds what keeps costing you points: the phase of the game where your accuracy drops, the tactical themes you miss, whether your errors are tactical or positional, and where you get short of time. Then it builds training from those weaknesses.

Prepare for a specific opponent

Give ChessMoves AI an opponent's name and it studies their games: their openings, how they answer your repertoire, where they score badly and where they make their mistakes. You can then play a whole game against a simulation of them. See AI chess opponent preparation.

Train and play

Rated puzzles, puzzle races, lessons built from your own mistakes, opening drills and visualisation training are in chess training. You can also play chess online against people or the computer.