What Is a Good Chess Accuracy? How the Number Is Made — and When It Lies

A good chess accuracy does not exist as a single number, and any chart claiming "1500s average 85%" is selling precision the metric cannot support. Accuracy measures the game as much as the player: the same person scores very differently in a short forcing win and a long defensive grind. Reading yours starts with knowing how the number is made.

What does chess accuracy actually measure?

Accuracy measures how much winning chance your moves preserved, compared against what a strong engine thought was available — it is a summary of evaluation swings, not a count of "best moves." The honest construction, and the one AxiomChess uses, works like this: Stockfish 18 evaluates every position of your game in one bounded pass — depth 24, capped at 8 million nodes — each evaluation converts to a winning-chance percentage, and every move is scored by how much of that winning chance it kept or gave away. Small concessions barely register; a move that drops your winning chances from 75% to 40% dominates the average.

Two properties follow from that construction, and both matter for reading your own number. First, accuracy is engine-relative: it grades agreement with one engine at one depth, so the tool should tell you which engine and how deep — ours is disclosed because a score you can't reproduce is a score you can't trust. Second, accuracy is side-honest: your score is computed from your own decisions, so a wild game where both players missed wins grades both players down — it does not care who eventually won. Scores from different sites are a separate question — different engines, depths, and formulas make them non-comparable, as covered in the game review guide.

What is a good accuracy in chess?

There is no universal benchmark, because the number is dominated by what kind of game you played. Three properties of a game inflate accuracy for both players, and none of them is playing strength:

• Forced sequences. Recaptures, only-moves, and known opening theory are moves

almost anyone finds — and in the winning-chance construction above they cost nearly nothing, so a game that is mostly forced grades everyone like a genius.

• Short games. Fewer real decisions means one clean stretch carries the score.

An opponent who collapses in 20 moves gifts you a spectacular number.

• Quiet, simplified positions. When many moves are near-equally fine, any

reasonable choice matches the engine closely. Sharp middlegames, where one precise path exists, punish everyone's percentage.

So the same 85% is excellent in a long, sharp, contested game and unremarkable in a short tactical collapse. If you want a usable reference point, build a personal one: compare accuracy only across your own games of similar length and type. Your average in long rapid games is a real baseline; a stranger's number from different games is not.

Why is your accuracy sometimes higher in games you lose?

Because accuracy grades decision quality, not outcomes — and lost games are often full of easy decisions. A player defending a worse position frequently faces forced replies for twenty moves and matches the engine on every one of them, while the winner navigates genuinely open choices and drops percentage doing it. This is also why AxiomChess labels every evaluation from the side to move: your score is built from your own decisions, never inferred from the result. High accuracy in a loss usually means the game gave you little room to go wrong, not that you were robbed.

The reverse also happens: you win a chaotic game where both sides traded mistakes, and the accuracy looks embarrassing. That is the metric working correctly. The result of one game is noise around the quality of its decisions — which is why "accuracy up, rating flat" is such a common and misleading combination. A number that tracked winners instead of decisions would feel better and measure less.

Should you try to raise your chess accuracy?

No — not directly. Accuracy is an outcome you read, not a lever you pull; chasing it optimizes the scoreboard instead of the play. The lever underneath is whichever mistake you repeat: unanswered threats (the blunder guide covers that habit), clock collapses (the time management guide shows how to read them), or winning positions that slip away. Remove a recurring leak and accuracy rises on its own, as a side effect of better decisions.

This is why the report card in AxiomChess is not an accuracy tracker: it grades your games against fifteen specific, checkable fundamentals — did you answer the threat, develop before attacking, keep your king safe — using deterministic rules, so every grade traces to a concrete position rather than a vibe. A percentage tells you that something leaked winning chances; a fundamentals report tells you what, and what to drill. The first is a thermometer, the second is a diagnosis.

FAQ

Is 90% accuracy good in chess?

Context decides. A short game full of recaptures and known theory can hand 90% to almost anyone; in a long, sharp, contested game the same number is genuinely strong. The number only becomes meaningful next to the game's length and how forcing it was — which is why no serious tool congratulates you on the percentage alone.

Is accuracy a good measure of chess strength?

It is a weak one. Accuracy varies more with game type than with the player, and outcomes of single games barely constrain it. Strength shows up in the quality of repeated decisions across many games — which specific fundamentals you keep and break — not in one engine-agreement percentage.

What accuracy should you aim for at your rating?

Don't aim at accuracy — no defensible per-rating table exists, because the number depends on game type. Aim at the leak the number hides: pick your most common recurring mistake, drill it, and let the percentage drift up as a consequence. Targets belong on habits you control, not metrics you don't.

Why does your accuracy drop in longer games?

More real decisions, fewer forced moves. A long game visits quiet positions where many moves are playable but one is precise, and every small concession compounds into the average. That drop is information, not failure — it marks the phases where your genuine decisions live, and where review time pays most.

Your accuracy number already has a story behind it — the specific decisions that made it. Analyze your games free — AxiomChess shows the winning-chance cost of every move and grades the fundamentals underneath the percentage.