What a free chess game review actually tells you

What free review already gives you

A free computer analysis on Lichess gives you a full engine pass over every move in the game: an evaluation graph, a label on each move (inaccuracy, mistake, blunder), the engine's preferred move in each flagged position, and the name of the opening. That is the substance of what a game review is. Nothing behind a paywall changes the evaluation of your position on move 23.

Chess.com's Game Review is the same idea with better presentation and a cap on how many you can run as a free account. That allowance has changed more than once, so check what your account currently gets rather than trusting a number in an article. Either way, the free analysis board on both sites lets you move pieces around any position with an engine running, which is the part most players skip.

So the honest answer to "is free chess game review good enough" is yes, comfortably, for anyone under about 1800. The limit is not the engine. The limit is that a list of flagged moves is not a lesson, and turning one into the other is work that no free review does for you.

What usually gets withheld

The paid tiers mostly sell you convenience, presentation and volume. Useful things, but rank them honestly:

• Unlimited reviews instead of a daily cap. Real value if you play a lot of games.
• Deeper search, sometimes a stronger cloud engine. At your level this changes the label on a handful of borderline moves and almost never changes the verdict on the moves that decided the game.
• Move-by-move commentary in words. Genuinely helpful when it is right, and it is often generic when it is not.
• Opening reports, drills built from your errors, aggregated stats across games.
• An accuracy percentage, which is a derived score, computed differently on different sites, and not comparable between them.

One move in 13: how much there is actually to review

Across 146,427 engine-classified moves in 4,684 analysed games on Mated, 3.8% of moves played are outright blunders and another 4.2% are mistakes. That is roughly one move in 13 being a real error, before you count inaccuracies, which run at 18.4%.

Do the arithmetic on a single game. If you made 40 moves, expect around three real errors. That is your review. Not twelve items, not a full-game walkthrough. Three moves, and probably one of them decided the result.

The same set of moves shows the errors are not spread evenly: 9.0% of middlegame moves are a mistake or worse, against 2.3% in the endgame. If you have ten minutes, spend them on the moves after your opening knowledge ran out and before the position simplified. That is where the damage happens.

There is a second number worth knowing. The average move in that set gives away 95 centipawns against the engine's best, but the median is far lower, because a small number of very bad moves carry most of the loss. Your play is not uniformly 95 centipawns off. It is mostly fine with occasional disasters. Reviewing the mostly-fine moves is how people spend an hour and learn nothing.

Reviewing a game with nothing but a board

Do this before you open any engine, on the game you lost most recently. It takes about eight minutes and it works because it forces you to commit to a judgement you can then be wrong about.

Then, and only then, run the free analysis. You are now checking a prediction instead of reading a report. The moves where your guess and the engine agree are moves you already understand. The moves where they disagree are the actual content.

• Replay the game from move one at normal speed, no stopping. Get the shape of it back.
• Find the last position where you would happily take your side. Write down the move number.
• Find the first position where you clearly would not. Write down that move number too.
• Look only at the moves between those two numbers. Usually there are three or four.
• For each one, say out loud what you were trying to do. "Defend b6." "Stop his knight coming to d5." If you cannot answer, that is the answer.
• Guess which single move was the error, and guess what you should have played instead. Commit to it in writing.

The flagged move and the losing move are often different

Here is the pattern that free review handles badly. Say you are Black, your knight sits on f6 in front of a castled king, and on move 19 you play Nd7 to shore up a pawn on b6. Three moves later Bxh7+ works, you take, and the attack lands. The engine puts a blunder tag on 19...Nd7 and offers a better move. Fine. But it does not tell you the thing you needed: your knight on f6 was the only piece covering h7, and you traded a defender for a pawn.

That is the general shape. The engine reports where the number moved. It does not report why your reasoning permitted it. Almost every error in this range comes from a short list — you removed a defender, you counted a capture sequence wrong, you moved on a plan while your opponent had a threat, you traded into a worse structure. Naming which one it was is the review. Reading the tag is not.

So when you look at a flagged move, ask what feature of the position you were not looking at. Then find one earlier move in the same game where you were not looking at it either. There usually is one.

When a tool is the answer and when it isn't

If you play two games a week, you do not need a product. Free analysis plus the eight-minute routine above covers it, and the routine is doing most of the work.

The case for a tool starts when you have volume. Thirty games a month is ninety-odd real errors, and no human is going to notice by eye that eleven of them were the same removed-defender mistake in the same pawn structure. That aggregation across games is the part free per-game review genuinely does not do. Mated exists for that: it pulls your recent games through the public Chess.com and Lichess APIs, runs Stockfish over every position in your browser, sorts your errors into five categories, and builds a fifteen-minute session from the positions you actually got wrong. Whether you want that or your own spreadsheet is a matter of taste.

What no tool can do is make you predict before you check. If you skip that step, an engine report is just a list of moves you did not play.

What nobody has measured properly

There is no good public evidence on how much game review improves rating, at any level, and anyone quoting you a figure has made it up. We can measure where your errors are; we cannot yet measure how much reviewing them moves your results, because that would need controlled comparison over months, and nobody has run it.

What is measurable and worth watching is your own error rate by phase. Take your middlegame mistake-or-worse rate now, and look at it again in fifty games. If it moves, something you are doing works. If it doesn't, and your rating moved anyway, you got better at something else — probably time management or opening familiarity — and it is worth knowing which.