Chess Game Review, Explained: What the Numbers Tell You — and What They Hide

A chess game review replays your finished game through an engine, labels each move, and hands you an accuracy score. Useful — and easy to misread. This guide explains what those numbers really measure, how to run a review that changes your next game, and which free tools are worth your time.

What is a chess game review?

A chess game review is an engine-checked replay of a finished game: software — almost always the Stockfish engine — evaluates every position you reached, flags the moments where the evaluation swung, labels your moves on a scale from brilliant to blunder, and summarizes the whole game as an accuracy percentage.

The format was popularized by the big playing sites and is now everywhere, from full-featured platforms to independent one-developer sites. The mechanics are broadly the same in all of them; what differs — a lot — is how deep the engine looks, how the labels and accuracy are calculated, and how honest the presentation is about both. That difference is the part most players never see, and it's where reviews mislead.

What do the game review numbers actually mean?

The numbers are the engine's judgment of best play from each position at a specific search depth — they are honest about the position, not about what was findable by a human, and not comparable from one site to another. Three things to internalize:

• Depth changes verdicts. A shallow search calls a move fine; a deeper search

calls it an error — same move, same position. Any review that doesn't disclose its depth is asking you to trust a number with no units.

• Accuracy scores don't transfer between tools. Each site converts evaluation

swings into labels and accuracy with its own formula and its own engine settings. The same game can score noticeably differently on two platforms without either being "wrong." Compare your accuracy within one tool over time, never across tools.

• Evaluations have a point of view. "+1.5" is only meaningful once you know whose

side it describes and whether it's centipawns or a winning-chances percentage. Winning chances are derived from the centipawn number, but they compress big material leads, which is why a "+8" and a "+12" feel identical in practice.

This is why disclosure matters more than polish. In AxiomChess, every position in every game gets the same bounded Stockfish 18 pass to depth 24, for both players, and evals are shown as centipawns and winning chances together — a number with its method attached is information; a number without one is marketing.

How do you review a chess game properly?

Review the game once from memory before the engine sees it — then use the engine to check your story, not to write it. The workflow:

1. Before the engine: mark the three moments you believe decided the game, and

write one sentence about each. This is the step that turns review into training — it tests your evaluation, not the engine's.

1. Engine pass: find the biggest evaluation swings and compare them against your

three moments. The mismatches are your blind spots, and they are the most valuable output of the whole exercise.

1. Contest the verdicts. Skip lessons from positions that were already lost, and

from engine lines you would never find in a real game. The turning points worth studying are the ones that were contestable at your level.

1. Name the cause, not the move. "I stopped checking his threats after I won the

exchange" is a lesson. "23...Rd8 was an inaccuracy" is trivia.

1. Log the cause and move on. One game deserves twenty minutes, not two hours —

the compounding value is in the log, not in any single game.

Which free chess game review tools should you know?

Judge review tools on three things: what they cost, whether they disclose their engine settings, and whether they help you act on what they found. The honest landscape, verified at the time of writing:

• [Lichess](https://lichess.org/analysis) — free and open source, with a full

analysis board, engine evaluation, and opening explorer. The default answer for no-cost, no-account analysis.

• [Chess.com Game Review](https://www.chess.com/analysis) — the most polished

guided review in the format, tied to a Chess.com account, with usage depending on your plan.

• [Chessigma](https://www.chessigma.com) — an independent free analysis site

offering game reports without a paywall.

• [WintrChess](https://wintrchess.com) — a free, open-source game analysis tool

that grew out of the earlier Game Report project.

• [AxiomChess](/) — full disclosure: this one is mine. It is a free

chess game analysis app that runs Stockfish 18 locally in your browser (your games never leave your machine), discloses its depth, reviews both players, and feeds every game into a cross-game report card of the fundamentals you follow and skip. If your games live in files, you can import PGNs directly.

Any of the first four will review a game competently. The differences that matter are disclosure and what happens after the review — which brings us to the real limit of the format.

Why doesn't a single game review make you better?

Because one game is an anecdote, and your rating is a pattern. A game review tells you what happened on Tuesday; improvement comes from noticing that Tuesday's mistake also happened the previous Thursday, and twice last month. The single-game format — however deep the engine — structurally cannot show you that.

The fix is boring and effective: keep the cause-log from step five, and read it every ten games. If you want the diagnosis automated, that cross-game layer is the entire reason AxiomChess exists. And if the pattern you find is that you're stuck in the same place month after month, the companion post on why you're not improving at chess covers how to turn a recurring cause into a training plan.

FAQ

Is chess game review free?

Free options cover most needs. Lichess offers free, open-source analysis; AxiomChess is free and runs entirely in your browser; several independent sites review games without a paywall. Some platforms tie deeper or more frequent reviews to paid plans, but nobody needs to pay to review their games well.

Are accuracy scores comparable between chess sites?

No. Each platform uses its own engine depth, its own move-labeling thresholds, and its own accuracy formula, so the same game produces different numbers on different sites. Neither number is a lie — they're different measurements. Track your accuracy within one tool over time and ignore cross-site comparisons.

Do you need the engine turned on to review a chess game?

Not at first — and turning it on too early is the most common review mistake. Do one pass from memory, committing to what you think the key moments were, then let the engine grade your story. Engine-first review turns you into a spectator of your own game.

How many games should you review each week?

Fewer than you think, reviewed properly. A handful of rapid or slower games with cause-tags beats reviewing every blitz game superficially. The goal is a growing log of named causes, and even two well-reviewed games a week builds one faster than twenty engine-skims.

A game review is a measurement, and measurements are only as good as their honesty. Analyze your games free — Stockfish 18 at a disclosed depth, both sides of every game, and a report card that shows you what keeps repeating.