Chess accuracy scores explained (and why 100% is not the goal)

Accuracy percentages are everywhere: 92%, 78%, 99%. They are easy to compare and easy to misunderstand. An accuracy score is a compressed summary of evaluation loss, not a measure of chess ability.

How the score is built

For each move, the engine compares the position before the move with the position after it. The difference in win probability becomes a loss for the player who moved. Those losses are averaged into a single percentage. A move that matches the engine's choice loses nothing; a blunder loses a lot.

ChessUrge uses the same win-probability model for its full-game reviews. Each move gets an accuracy value, and the game summary averages them for White and Black.

Why short games inflate accuracy

A 12-move miniature with a forced opening line may have very few real decisions. If both players follow theory and then one blunders, the winner can finish with 98% accuracy. A 60-move grinder with several difficult endgame choices may produce 82% for both players despite excellent play. The percentage reflects the position, not the difficulty of the decisions.

What to look for instead

• Where did the big losses happen? Opening, middlegame or endgame?
• Were they tactical or positional? A missed fork is a different problem from a bad pawn break.
• Did time pressure matter? Check the move numbers and your clock.
• Was the best move findable? A 12-move engine line is not a human plan.

Accuracy is a diagnostic, not a grade

Do not chase 100%. Chase a smaller number of repeated mistakes. If your accuracy is 85% but your three worst moves are all the same kind of missed tactic, you have a clear training target. If your accuracy is 95% but you lost on time, accuracy is not the story.

Use the number to find the review, then put the number away. The move list and the position are what matter.

Run a full review on your game library and compare the accuracy breakdown with the move markers.