What accuracy actually measures
After a game review you get an accuracy score, usually from 0 to 100%. It answers one question: how close were your moves to the engine's best play? A move that matches Stockfish scores near 100%; a move that throws away a winning position scores low. Your game accuracy is a weighted summary of all those individual move scores.
How it is calculated
The engine evaluates the position before and after each of your moves. The size of the evaluation drop — how much worse your position got compared to the best move — is converted into a per-move score. Small, forgivable inaccuracies barely dent it; a blunder that loses a piece drops it sharply. Those per-move scores are then combined into one game figure. Because the drop matters more than the raw evaluation, one player can have high accuracy in a wild game and another low accuracy in a calm one.
What is a good accuracy?
- Under 1000: 60–75% is normal — expect a few blunders per game.
- 1000–1600: roughly 75–85%.
- 1600–2000: often 85–92%.
- Masters: frequently 95%+ in quiet games.
Do not chase 100%. A sharp, complicated game with an 85% can be far better chess than a lifeless 95% draw.
How to use it to improve
Accuracy is a headline, not a lesson. The value is in where you lost points. Open your game analysis, find the moves that cost the most, understand the better move, and then drill those positions with blunder puzzles. For the labels behind the number, read Blunder, Mistake or Inaccuracy?