Chess Move Classifications Explained: Blunder, Mistake, Inaccuracy, and Best

What the move grades in a game review actually mean, how engines decide them, and why a "good" move in one position is a blunder in another.

Open any modern game review and every move you played gets a label — Best, Good, Inaccuracy, Mistake, or Blunder. These classifications are the fastest way to see where a game slipped away. But what do they actually measure, and why does the same move get graded differently depending on the position? Here is how it works under the hood.

The core idea: how much did the move cost?

Every classification comes from one question: how much worse is the position after your move compared to the best available move? An engine like Stockfish evaluates the position before and after you moved. The gap between the best move’s evaluation and your move’s evaluation is the “cost” of your decision.

Older tools measured this cost in centipawns (one centipawn = one hundredth of a pawn). If the best move kept you at +0.3 and you dropped to −0.9, that is a swing of about 120 centipawns. The bigger the swing, the worse the label.

Why centipawns alone are misleading

Raw centipawn loss has a famous problem: it punishes you the same whether you are winning easily or fighting for your life. Going from +9 to +6 is a 300-centipawn drop, but you are still completely winning — that is not a blunder in any practical sense. Going from +0.5 to −2.5 is the same numeric drop, but it just threw away the game.

That is why better tools — including Chesslume — convert the evaluation into a win probability first, then measure how much your winning chancesdropped. A move is judged by how much it actually changed your odds of winning, not by an abstract pawn count. This is why a sloppy move in a totally winning position is correctly graded as merely “Good” instead of flagged as a disaster.

The grades, from best to worst

• Best— you found the engine’s top choice (or something equal to it). Your winning chances barely moved.
• Good— not the engine’s first pick, but it keeps your position healthy. The cost is small.
• Inaccuracy — a noticeable slip. You gave away some of your advantage or made your position harder than it needed to be.
• Mistake — a real error that meaningfully shifted the evaluation against you. These are the moves worth studying.
• Blunder — a move that swung the game. A winning position becomes equal, or an equal position becomes lost.

Some platforms add extra tiers like Brilliant (a strong move that involves a sound sacrifice) and Great (the only move that holds the position), plus a Book tag for known opening theory. These are presentation layers on top of the same underlying win-probability math.

A worked example

Say you are up a rook and playing at +6.2. You play a move that lets your opponent trade queens for no compensation, and the evaluation drops to +4.8. In raw centipawns that is a swing of 140 — enough to be flagged as a Mistake by an old-style tool. But your win probability barely moved: you were roughly 99% to win before and after, since a two-pawn lead with queens off the board is still completely winning. A win-probability grader correctly calls this Good or, at worst, an Inaccuracy, because nothing about your actual chances changed.

Now compare a quiet middlegame at +0.4. You hang a pawn for nothing and drop to +0.1 — only a 30-centipawn swing, tiny by raw-centipawn standards, but in a balanced position that can shift your win probability from 55% down to 51%, a real if minor slip. Push it further: a move in a roughly equal position that collapses into a lost endgame might only move the raw evaluation from 0.0 to −3.0 (a 300-centipawn swing) while your win probability falls by 80 percentage points or more. Similar-looking centipawn numbers, completely different verdicts, because what matters is what the move did to your practical chances, not the number on its own.

Two common misconceptions

First, an Inaccuracy is not a small Mistake — on its own it rarely costs a game, so treat a single one as noise. Three or four Inaccuracies of the same type across your recent games is the real signal worth acting on. Second, a Blunder does not always mean you lose. Your opponent still has to find the punishing reply, and at club level they often miss it. That is exactly why the report is more useful than the final result: the label tells you what should have happened, independent of how the game actually ended.

Why context changes everything

The single most important thing to understand: the same move can be Best in one position and a Blunder in another. A queen trade is brilliant when you are up material and want to simplify, and a blunder when it walks into a back-rank mate. Classifications are always relative to the specific position, which is exactly why an engine has to recalculate from scratch on every move.

How to actually use them

Do not obsess over every Inaccuracy. The highest-value habit is to jump straight to your Mistakes and Blunders, set up the position, and ask yourself what you missed before looking at the engine line. That is where rating points hide. Once you spot the pattern in your own blunders — hanging pieces, missed tactics, time-pressure collapses — you know exactly what to train.

You can see all of this on your own games in seconds. Import a game into Chesslumeand every move is graded with the win-probability method described above, with the engine’s best line one click away. Want to drill the Mistakes and Blunders pattern on demand instead of waiting for your next game? Try the puzzle trainer.