What Does Accuracy Mean in Chess? How It's Calculated and Used

Updated November 2026 · 8 min read · Every position checked with Stockfish

Every game review on Chess.com and Lichess shows an accuracy percentage, and most players either obsess over it or ignore it. Both are mistakes. This guide explains what the number actually measures, how each site calculates it, how it connects to inaccuracies, mistakes and blunders, why it doesn't map neatly onto rating, and how to use it so your rating goes up.

1. Understand what accuracy actually measures

Accuracy compares every move you played with what a strong engine, usually Stockfish, considers best in that position. If your move keeps your chances about the same as the engine's choice, it scores near 100. If it throws those chances away, it scores low.

The key detail is that both sites measure the loss in winning chances, not raw pawns. The engine's evaluation is converted into a probability of winning, and each move is judged by how much that probability fell.

That has two consequences:

• When the position is balanced, a small error can drop your chances a lot, so it costs accuracy.
• When you are already completely winning (or completely lost), even a sloppy move barely changes your chances, so it costs little.

Accuracy is therefore a measure of how much you gave away, move by move, relative to perfect play. It says nothing directly about how difficult the position was or how well your opponent played.

2. How Chess.com calculates accuracy

Chess.com's accuracy is based on an expected points model: the engine evaluation is turned into the points you'd expect to score from the position, and each move is scored by how many expected points it lost. Those move scores are combined into one number for the game. Chess.com explains the idea in its support article but hasn't published the full formula in the same detail as Lichess.

Alongside the percentage, Game Review labels individual moves: Brilliant, Great, Best, Excellent, Good, Book, Inaccuracy, Mistake, Miss and Blunder. A Miss means your opponent made an error and you didn't punish it.

Game Review can also show an estimated game rating for each player. People often search for this as the "chess.com accuracy rating". It is a rough performance estimate for that single game, based on how you played, and it is not your actual rating. Treat it as a curiosity, not a measurement. For a breakdown of what typical scores look like, see what a good accuracy on Chess.com is.

3. How Lichess calculates accuracy

Lichess publishes its method. In short:

• Each engine evaluation is converted into a win percentage for the player to move.
• Each move's accuracy is calculated from how much your win percentage dropped. A tiny drop scores close to 100. A large drop pulls the score down steeply.
• The game's accuracy combines two averages of those move scores: one weighted by how volatile the position was, and a harmonic mean. The harmonic mean is deliberately harsh on very low scores, so a single collapse drags the whole game down more than a simple average would.

Because the two sites use different models, the same game can get a different accuracy on each. That is expected and doesn't mean either is wrong. Compare your accuracy only with your own games on the same site and in the same time control.

4. Know how bad inaccuracies, mistakes and blunders really are

Inaccuracy, mistake and blunder are bands of the same thing: how much a move cost you. An inaccuracy gives away a little, a mistake gives away a clear amount, and a blunder) usually changes the result of the game. The exact thresholds differ between sites.

Are inaccuracies bad? They are imperfect, but at 600–1800 they rarely decide games. A handful of inaccuracies with no blunders is usually a well-played game. One blunder can wipe out forty good moves.

Here is the classic example of a move that ruins a game in one go:

The position before it is ordinary. Black has reasonable defences such as g6 or Qe7. Instead, Nf6 allows Qxf7 checkmate. In accuracy terms, that single move takes Black's winning chances from roughly even to zero.

The practical rule: count your blunders and mistakes first and inaccuracies last. If you keep losing material in one move, start with how to stop hanging pieces.

5. Don't treat accuracy as a rating

Your rating comes only from results. Chess.com and Lichess both use Glicko-based systems, which update your rating based on who you played and whether you won, drew or lost. Accuracy has no direct effect on it.

The two are related on average, because stronger players give away less. But a single game's accuracy is a noisy signal of strength:

- Short games inflate scores. A miniature where your opponent blunders early gives the winner a very high number from very few moves.

So the answer to "chess accuracy vs rating" is this: averaged over many games, accuracy tracks strength loosely. In any one game it can easily mislead you.

6. Use accuracy to find what to fix

Accuracy is most useful as a pointer to specific moves. A simple routine after each game:

• Open the review and go straight to the largest drop in the evaluation graph. That one move usually explains the result.
• Before you check the engine line, ask yourself what you missed. Was it a hanging piece, a check, or a threat you didn't see?
• Note the type of error and the move number. After a couple of weeks, patterns appear: the same opening, the same phase, the same kind of tactic.
• Track your average accuracy and blunder count over many games in one time control, not single scores.

The full method is in how to analyse your chess games.

How Chessy helps: connect your Chess.com or Lichess username and Chessy imports and analyses your games with Stockfish. It charts accuracy over time and your blunder rate, and shows the move numbers and times of day where your results drop. It also turns your own blunders into puzzles, so you drill the exact positions you got wrong. See why puzzles from your own games work better.

Common mistakes

• Comparing accuracy across sites or time controls. Chess.com and Lichess use different formulas, and blitz produces lower scores than rapid. Comparing a Lichess blitz score with a Chess.com rapid score tells you nothing. Compare like with like.
• Chasing a high number instead of wins. Choosing dull, safe positions can raise your accuracy and lower your results. Play good moves and let the score follow. Results are what move your rating.
• Trusting a high score from a short win. If your opponent blundered on move eight, your high accuracy says more about them than about you. Long, hard-fought games are a better test.
• Reading the percentage and skipping the moves. The number on its own teaches you nothing. The value is in the one or two moves where your chances collapsed and in understanding why you played them.
• Worrying about every inaccuracy. Fixing small inaccuracies while you still hang a piece every few games is the wrong order. Cut blunders first. Refine later.

FAQ

Accuracy is a summary of how much you gave away, not a verdict on how good you are. Stop reading it as a grade and start using it as a map to the one or two moves that decided each game. Fix those patterns, and your rating, which only counts results, will follow.

Sources

• Chess.com Support: What is Accuracy? — Chess.com's expected points approach to accuracy
• Lichess: how accuracy is computed — Win percentage conversion, per-move accuracy and the harmonic and volatility-weighted averages
• Wikipedia: Blunder (chess) — Definition of a blunder
• Wikipedia: Glicko rating system — How ratings are updated from results, not accuracy
• Chess Programming Wiki: Stockfish — Background on the engine used for evaluations