}

Chess, with clarity.


Zero-Training AIDeterministic chess engine
BAYESIAN Untimed
0.00
EVAL
Y
YouPlaying as White
Your move
Select a piece, then a highlighted square.↑↓←→ + Enter
ZeroTrainingAIChess/Built to make the thinking visible.Local C# engine·No cloud calls·
A NEW PERSPECTIVE

Make your move.

A clean board. A fresh line of thought.

Choose your side
Start from a custom position

Basic legality is checked. A FEN does not contain earlier repetition history.

A fresh game starts immediately. Refreshing this page also resets Chess to the standard starting position.

A well-earned promotion.

Choose your new piece.
UNDER THE SURFACE

Nothing hidden.
Just chess.

The playing engine

A new C# reference implementation uses chess rules, hand-written evaluation features, iterative-deepening alpha-beta search, and capture quiescence. It is not Stockfish, an LLM, or an integration of a proprietary Zero-Training AI SDK.

Always-on Bayesian Optimization

Bayesian Optimization is permanently active. A Gaussian-process surrogate learns from locally generated benchmark observations and selects new evaluation-weight candidates using expected improvement. The best verified weight set is applied automatically to normal engine play. There is no switch and no manual apply step.

What the numbers mean

Positive evaluations favor White; 100 centipawns equal one pawn unit. Search depth is the last fully completed iteration; tactical extensions may look further. Fixed node budgets are reproducible. Tuning scores are limited benchmark proxies—not Elo, proven strength, or a win guarantee.

Playing and privacy

Click a piece and its destination, or use arrow keys and Enter. Castling, en passant and all four promotions are supported. Threefold repetition and the 50-move rule are claimed automatically in this casual app. Basic insufficient-material draws are detected; arbitrary locked dead positions are not exhaustively recognized. No clocks or tournament adjudication. The current game exists only while this page is open; refreshing starts a fresh game. The always-on Bayesian optimizer runs in the local C# process.

N New gameU Undo turnH HintF Flip board