Chess Game Report Explained: What Your Accuracy Score Really Means

Your accuracy percentage, centipawn loss and move classifications decoded — so you can stop chasing numbers and start fixing the right things.

After Stockfish finishes analyzing a game, you get a report: an accuracy percentage for each side, a count of best moves, inaccuracies, mistakes and blunders, and an overall centipawn loss. Most players glance at their accuracy and move on. This guide explains what every number actually measures — and which ones are worth paying attention to.

What the accuracy percentage measures

Accuracy converts the average centipawn loss per move into a win-probability scale and expresses it as a percentage. A score of 100% means every move was the engine's top choice; a score of 50% means you consistently played moves that halved your winning chances. The conversion is not linear — one blunder on move 10 can drop your accuracy by 15 points even if every other move was perfect.

Centipawn loss: the raw signal

One centipawn (cp) equals one hundredth of a pawn. If you play a move that costs you the equivalent of half a pawn, that is 50 centipawns. Average centipawn loss (ACPL) is the mean across all moves. Grandmasters average below 20 ACPL in classical games; club players typically see 50–120 ACPL depending on time control. Blitz naturally shows higher ACPL because there is less time to find the best move.

Move classification labels

• Brilliant (!!) — a strong, non-obvious sacrifice or defensive resource that the engine rates highly. Rare and genuinely impressive.
• Best move (!) — the engine's top-ranked move. Consistently finding best moves is the clearest marker of improvement.
• Good move — a solid move within a small margin of the best.
• Inaccuracy (?!) — loses roughly 0–1.5 pawns of advantage. Suboptimal but not immediately punishing.
• Mistake (?) — loses 1.5–3 pawns. A serious error that changes the character of the position.
• Blunder (??) — loses more than 3 pawns or allows a forced win for the opponent. Game-changing.

How to use the report correctly

Do not read the report as a final grade — read it as a map. The move classifications tell you where in the game things went wrong. Jump straight to your blunders and mistakes, not to your best moves. For each error, ask one question: what was my opponent threatening that I missed, or what did I think the position required that the engine disagreed with?

Critical moments matter more than the average

A single blunder in a completely equal endgame is more instructive than five inaccuracies in a position that was already lost. When reviewing, pay attention to the shape of the evaluation graph: a sudden drop shows a critical moment; a gradual decline shows a strategically mishandled position. Both patterns require different fixes.

Get a full game report on any of your games — free, no sign-up.