ChessGrader / Guides / What is a good accuracy?
What is a good accuracy in chess?
A good accuracy depends on your rating and the time control — there is no single number. As a headline range: 70–80% is solid for club players, 80–90% is strong, and 90%+ in a slow game is master territory.
That's the short answer. For most players rated 1000–1600, anything in the 70s is a perfectly good game, and a game in the 80s usually means you played well or your opponent made it easy. Below is the longer answer: what typical accuracy looks like at each rating band, why the number swings so much from game to game, and why chasing it is mostly a waste of your time.
Want the number for a specific game? Use the free chess accuracy calculator with a Chess.com username, Lichess username, or PGN.
Typical accuracy by rating
These are practical ranges synthesized from published game samples, site explanations, and community-reported results. ChessGrader has not measured a large enough independent rating-stratified dataset to call them population averages. Treat them as guideposts for rapid-to-classical games, not hard cutoffs. In bullet, everyone's numbers usually drop.
| Rating | Typical accuracy | What it looks like |
|---|---|---|
| Under 1000 | ~55–70% | Several mistakes per game, occasional hung pieces |
| 1000–1600 | ~70–80% | Fewer blunders, but critical moments still go wrong |
| 1600–2000 | ~80–90% | Clean games with one or two real errors |
| Masters / GMs | 90%+ in slower games | Near-engine play in quiet positions |
One game above or below your band means nothing. A trend across twenty games means something.
Why accuracy swings wildly from game to game
Accuracy measures how closely your moves matched the engine's, with each deviation scored by how much it changed the position. That makes it extremely sensitive to what kind of game you played:
- Forced sequences inflate it. If your opponent walks into a line where every one of your moves is a forced recapture or an obvious check, you rack up "best move" after "best move" without doing anything hard.
- One-sided games inflate it. When you're up a queen, almost everything wins, so almost nothing you play loses meaningful win probability. Crushing a much weaker opponent routinely produces 90%+ accuracy that says little about your strength.
- Sharp, balanced games deflate it. A long, tense struggle against an equal opponent has many moments where only one or two moves hold. Even strong players score in the 70s in games like that — and those are often their best games.
A 95% in a 12-move miniature and a 78% in a 60-move grind are not comparable numbers. The grind was probably the better game.
How accuracy is actually computed
Every accuracy score starts with engine evaluations before and after each move. ChessGrader converts those evaluations through a fixed, rating-neutral expected-points curve reserved for accuracy, with explicit safeguards for missed mating sequences. Its accuracy score is a CAPS2-inspired approximation calibrated against reviewed games available through Chess.com's public archive. Per-move scores are combined with a school-test-style per-move curve and an ordinary mean. Book and Forced moves score 100; other moves lose credit with the accuracy-specific expected-points loss. This is separate from the rating-aware loss and centipawn safeguard used for move labels. It keeps one catastrophe important without letting it mathematically overwhelm thirty precise moves.
Chess.com's exact CAPS2 formula is private. ChessGrader documents its approximation, calibration target, thresholds, safeguards, and engine settings on the methodology page.
Why the same game scores differently on different sites
Paste one game into Chess.com, Lichess, and ChessGrader and you'll get three accuracy numbers that may differ noticeably. All three use Stockfish, but they differ on search depth, on how centipawn loss maps to a per-move score, and on how per-move scores combine into a game score. Chess.com's exact formula is private; Lichess publishes its method; ChessGrader publishes a CAPS2-inspired approximation calibrated against public reviewed-game benchmarks. Different scores do not mean any tool is broken—accuracy is a modeling choice, not a physical constant.
Why chasing the number is a trap
Accuracy is an output, not an input. You can't practice "being more accurate" — you can only fix the specific things that cost you win probability. The productive loop looks like this:
- Review the game and find the two or three moves that actually lost the most.
- Ask what you were thinking at each one — didn't see it, saw it and misjudged it, or rushed.
- Fix the pattern (tactics for the first, evaluation for the second, clock discipline for the third).
Do that for a few months and your accuracy rises on its own, as a side effect. Players who fixate on the score instead tend to play passively — trading everything off to avoid mistakes — which produces pretty numbers and stalled ratings. A structured game-analysis habit beats staring at the percentage.
Rule of thumb: track your blunder count per game, not your accuracy. Getting from two blunders per game to one will do more for your rating than any accuracy target.
Frequently asked questions
Is 80% accuracy good in chess?
For most players, yes. 80% is roughly the top of the typical range for club players rated 1000-1600 and normal for players rated 1600-2000 in rapid or classical games. Context matters though: 80% in a sharp, balanced game is a strong result, while 80% while crushing a much weaker opponent is unremarkable.
Why is my accuracy sometimes higher in bullet than in rapid?
Usually because bullet games are shorter and more one-sided. If your opponent blunders early, most of your remaining moves are easy and score well, inflating the average. Over a large sample, almost everyone is less accurate in faster time controls — single games are just noisy.
Does high accuracy mean I played like a grandmaster?
No. Grandmasters average 90%+ in slower games against strong opposition, which is much harder than scoring 90%+ in a game your opponent lost quickly. Accuracy measures how closely you matched the engine given the positions you faced; easy positions produce high scores at any skill level.
Can accuracy scores detect cheating?
Not on their own. A single high-accuracy game is normal and proves nothing, since forced sequences and one-sided games routinely produce 90%+ scores for honest players. Real cheat detection looks at statistical patterns across many games, move timing, and agreement with engine choices in hard positions. Never accuse someone based on one accuracy number.
What accuracy do grandmasters get?
Typically 90% or higher in classical games, sometimes approaching the high 90s in clean wins. In blitz and bullet their accuracy drops like everyone else, often into the 80s. Even world-class players score in the 70s occasionally in long, complicated fights.
Why does ChessGrader give me a different accuracy than chess.com?
Chess.com's exact CAPS2 formula and search settings are private, so no other tool can promise reproduction. ChessGrader uses a CAPS2-inspired approximation calibrated against public reviewed-game benchmarks and documents that method. Scores may differ materially, especially in short, one-sided, or tactically sharp games.
Does accuracy matter more than rating?
No. Rating measures results against real opponents, which is what actually matters. Accuracy is a useful diagnostic for individual games, but it is inflated by easy games and deflated by hard ones. Use it to find your worst moves, not as a scoreboard.