Brilliant, Best, Mistake, Blunder: move classifications explained

What each move label means in ChessAnalyzer, how it is calculated from winning chances, and why a Brilliant move is rare.

Every move in a ChessAnalyzer review gets a label. The labels are not opinions: each one comes from a rule applied to what Stockfish 18 thinks of the position before and after the move. This page explains every label and the rule behind it.

Winning chances, not pawns

Most engines show an evaluation in pawns: +1.50 means White is ahead by about one and a half pawns. But a pawn is worth very different amounts in different positions. Going from +0.50 to −1.00 changes the game; going from +8.00 to +6.50 changes nothing, because White is still completely winning. So ChessAnalyzer converts every evaluation into winning chances — a number from 0 to 100% — and judges moves by how many winning chances they give away. The conversion is the logistic formula used by Lichess, with the constant 0.00368208 per centipawn.

The labels

Book

A move from established opening theory, recognised in a database of more than 3,000 named opening positions.

Only move

The only legal move, or the only move that does not lose. There is nothing to judge.

Best

The engine's top choice, or a move just as good.

Excellent and Good

Moves that give away very little — a few percent of winning chances at most. Most moves in a well-played game land here.

Inaccuracy (?!)

A move that gives away at least 10% of winning chances. Not a disaster, but there was something better.

Mistake (?)

At least 20% of winning chances lost. These usually change the course of the game.

Blunder (??)

At least 30% lost — often a piece, or a winning position thrown away.

Miss

You had a forced mate or a clearly winning advantage (around +2.00 or more), and the move you played let it slip below a winning level. Misses are the opportunities you didn't see.

Great (!)

A move that was clearly the only good choice: the second-best alternative was much worse. Finding these under pressure is a sign of strong calculation.

Brilliant (!!)

A sound sacrifice. The move gives up at least one and a half pawns of material, is not just recapturing, and is still the best or nearly the best move — without simply being a winning position anyway. Brilliant moves are deliberately rare.

Why the thresholds are fair

Because labels use winning chances, the same rule works in the opening, the middlegame and the endgame, and in level positions as well as winning ones. A move that drops a pawn in a lost position will not be called a blunder, and a quiet move that turns a draw into a loss will be — which matches how strong players judge games.

A word of caution

Labels describe what the engine found at the depth it searched. Very deep, long-term ideas can occasionally be misjudged, and a "Mistake" in a practical game can be a perfectly reasonable human decision. Use the labels to find the moments worth studying, then think about them yourself. See also: how accuracy is calculated.