about this demo
We have trained a state-of-the-art machine learning model on intermediate-level chess player games. Think of it
as an LLM, but rather than predicting the next word, we predict the next chess move the player will make. The
model is run right in your browser, so it may take a moment to load. We don't need to send any moves back to the
server.
On top of the machine learning model, we have developed the CHESSMATE algorithm (Calculated HEuristic
Strategy and Statistical Modeling of Anticipated Tactical Engagements) which uses its predictions to play the
best move against the model. This algorithm can trick you into playing a bad move to beat you quickly. In other
words, Stockfish (a strong chess engine) will try to beat you in 70 moves, but our algorithm tries to beat you
in half the time with riskier moves.
The machine learning model should run in less than a second, but the CHESSMATE algorithm may take a few seconds
due to a high number of Stockfish calculations.
Model status: loading