Live evaluation
An evaluation bar and graph that update as you move through the game, with the numeric score on the winning side. Forced mates fill the bar outright.
Paste a game, drop in a FEN, or play out a line — and get instant NNUE evaluation with multiple candidate lines, a live evaluation graph, and a full move-by-move review that marks every mistake and estimates each player's rating. Runs in your browser, or install it and run it entirely on your own machine.
Free. No account. Your games stay on your device.
An evaluation bar and graph that update as you move through the game, with the numeric score on the winning side. Forced mates fill the bar outright.
MultiPV: see the top candidate moves side by side with their evaluations, click any line to play it out or insert it as a variation.
Reviews every move and labels it — Brilliant, Best, Excellent, Good, Book, Inaccuracy, Mistake, Blunder — then reports accuracy and an estimated Elo per side.
A proper variation tree with comments, NAG symbols and sub-lines — not a flat move list. Import and export standard PGN.
Import PGN databases and see what was actually played from any position, with win/draw/loss rates and move popularity.
Play the engine at calibrated strengths, run engine-vs-engine matches, drill openings with spaced repetition, and solve tactics mined from your own games.
Open it now — nothing to install. Analysis on the server is time-limited so it stays fast for everyone.
The full app with no search limits — deeper analysis, unlimited engine matches, and your database stored locally.
The default analysis engine. Stockfish evaluates positions with an efficiently-updatable neural network (NNUE), which is where the analysis strength and the NNUE evaluation come from. Open source, GPL-3.0.
Our own UCI engine, written entirely in Rush, the RiseOJ language — it transpiles to C++ and runs at native speed. Bitboard move generation (validated against published perft counts), alpha-beta search with a transposition table, and MultiPV. It also supports NNUE nets via EvalFile, though its classical evaluation is currently the stronger default.