Self-Learning Chess

A model that starts knowing nothing and learns from the games it plays — against you, or against itself when you train it.

Your move
Play as
What the model knows
Positions learned—
Games played—
You 0Drew 0Model 0
Train by self-play

Let it play itself and learn from every game — same learning as playing you, just thousands of times faster.

Self-play games0

Early on it plays almost random legal moves. Beat it (or fool it) the same way a few times — or run a batch of self-play — and watch “positions learned” climb. Its play shifts slowly, like a human beginner.