What Is a Good Accuracy Score in Chess?

You win a sharp game, run the review, and the report says 84. Last week you lost with a 91. A streamer you watch posts 96 in a blitz game and calls it an off day. Before any of those numbers can tell you something, you need to know what the score measures and what makes it move.

The short version: for most club players, anything above 80 in a normal-length game is a good result, and the low 90s is an excellent one. The long version is worth five minutes, because the same score can mean different things in different games, and chasing the number for its own sake will slow your improvement down.

What the Percentage Measures

Accuracy answers one question: how close did your moves stay to the engine's choices? Stockfish evaluates every position in your game, compares your move to its best line, and the gap on each move feeds a single score for each player.

Sites turn that comparison into a percentage in different ways. On Chess It Up, the review grades every move first, from Brilliant down to Blunder, and each grade carries a fixed value. Best moves sit near the top of the scale, Good moves take a real deduction, and a Blunder is worth almost nothing. Your accuracy is the average of those values across all your moves. One blunder in an otherwise clean forty-move game costs you about two points. Three blunders in a twenty-five-move game cost more than ten, and no amount of clean play around them buys the points back.

The design is deliberate. A Good move keeps your position healthy, and it still scores well below a Best move, because the gap between "fine" and "strongest" is where games at every level get decided.

Accuracy Benchmarks by Rating

The ranges below describe a typical game of thirty to forty moves at rapid or classical time controls. Treat them as rough guides. A single game can land far outside your band for reasons covered in the next section.

Rating	Typical accuracy range
Under 800	55–70%
800–1200	65–78%
1200–1600	72–83%
1600–2000	78–88%
2000+	83–92%
Titled players	90%+

Two things stand out in that table. The bands overlap, so a 1400 posting 80% has played a game a 1700 would be content with. And the ceiling rises slower than you might expect: masters lose accuracy too, just in smaller amounts and at worse moments for their opponents.

A more useful benchmark than any table is your own average. Pull your last ten games, note the mean, and treat any game above it as a good one. The player who moves from 74 to 79 over a month has made real progress, whatever the table says about 79.

Why Your Score Swings Twenty Points Between Games

Accuracy measures your moves against the position in front of you, and positions vary enormously in how hard they are to play. Four factors move the number more than your chess does.

Your opponent's strength. A weaker opponent hangs pieces and walks into tactics, and the punishing moves are easy to find. Strong opponents steer the game into positions where the best move is subtle, so your score drops even when your play improves. Beating up on lower-rated players inflates accuracy; losing well against higher-rated players deflates it.

How forcing the game is. Recaptures, forced trades, and only-moves all count toward your score, and the engine agrees with almost all of them. A game with three long forced sequences pads your accuracy with moves that required no thought. A quiet middlegame where eight candidate moves all look reasonable is where the score gets honest.

Game length. Each move carries equal weight, so length changes what a mistake costs. In a sixty-move endgame grind, one blunder disappears into the average. In an eighteen-move miniature, the same blunder is a twentieth of your score.

Time control. Blitz accuracy runs well below rapid accuracy for the same player, because pattern recognition alone finds fewer best moves than pattern recognition plus a think. Compare blitz games to blitz games.

A 95 Can Mean Less Than an 80

Take two games. In the first, your opponent trades every piece by move twenty-five and offers a draw in a dead-equal endgame. Half your moves were recaptures. You score 95.

In the second, you defend a piece sacrifice for thirty moves, find a string of only-moves under pressure, and lose anyway when a single defensive resource escapes you on move fifty-one. You score 80.

The second game is the better performance and by a wide margin. The score has no way to say so, because it averages over moves without weighing how hard they were to find. Read your accuracy the way you would read a test score without seeing the test: useful, and only meaningful once you know how difficult the questions were.

Don't Compare Scores Across Sites

Chess.com, Lichess, and Chess It Up each compute accuracy with their own formula. Chess.com fits a curve to win-percentage swings. Lichess derives its number from centipawn loss. Ours averages the move grades, as described above. The same game will produce three different percentages, and none of them is wrong. They are different rulers. Compare your Tuesday score to your Monday score on the same site, and ignore the rest.

The Numbers That Predict Improvement

Accuracy is a summary, and summaries hide the part you can fix. Two companion numbers do more for your rating than the headline percentage.

Average centipawn loss (ACPL) measures how much evaluation you gave away per move, in hundredths of a pawn. An ACPL under 25 is strong club play; under 50 is respectable; over 80 means large mistakes are frequent. ACPL responds faster than accuracy when your play changes, because it weighs mistakes by size instead of counting grades.

Blunder count is the blunt one. Below 1600, games are won by the player who blunders less, and a plain count per game tracks the skill that decides them. Cutting from three blunders a game to one is worth more rating than any opening you will ever learn.

Both numbers appear in every free game review alongside accuracy. Check them over your last twenty games rather than any single one; the trend is the signal, one game is noise. Our guide to analyzing your own games covers what to do with the moves the review flags.

Frequently Asked Questions

Is 90% accuracy suspicious?

Not on its own. Short games, forcing games, and games against opponents who collapse early produce scores in the 90s for players of any strength. Fair-play systems look for sustained engine-level play across many long, difficult games, which is a different pattern from one clean win. If you hit 94 in a game where your opponent hung a rook, you earned it.

Why did I score higher in a game I lost?

Accuracy grades your moves, and the scoreboard grades the result. Lose a long, clean game to someone who played cleaner, and your score will beat plenty of your wins. One blunder can decide a game in which you outplayed your opponent for forty moves. The mismatch is information: the loss came down to one moment, and the review will show you which.

Can I score 100%?

Not here. Our scale reserves the top of the range for Brilliant and Great moves, so a game of pure engine-best play lands at 96. Scores above that mean you found something the engine considers out of the ordinary.

Should I try to raise my accuracy directly?

Raise the inputs instead. Fewer blunders, fewer moves played on autopilot, a habit of checking your opponent's threats before your own ideas. Accuracy follows those changes on its own, usually with a lag of a few weeks. A player who optimizes the number itself starts avoiding sharp positions, and sharp positions are where you learn.

Find Your Baseline

The benchmarks above are averages over other people's games. Your own number is a ten-minute job: import your recent games, run the review on each, and note the accuracy, ACPL, and blunder count. That average is your real baseline. Come back in a month of honest play and measure again. The direction of that comparison matters more than any figure in the table.