How to review a chess game
Running the engine is the easy part. This is what to do with the result — and which of your mistakes are worth any of your attention.
Start at the graph, not at move one
A finished review gives you an evaluation graph. Most games are decided by two or three moments, and the graph shows you exactly where they were: the cliffs. Click the biggest one and you are at the position that decided the game, without stepping through thirty moves of nothing to find it.
Reviewing a game in order is how most people do it, and it is why most reviews stop at move fifteen out of boredom.
Ask what you saw, before you look at the answer
Sit on the position for a moment before revealing the engine's move. What were you intending? What did you think your opponent's threat was? The gap between what you thought was happening and what was actually happening is the thing worth learning. The engine's move on its own teaches you almost nothing — you will not be in this exact position again.
Sort your mistakes into three piles
- You missed something concrete. A hanging piece, a fork, a back-rank mate. These are the cheapest to fix: they are a checking habit, not knowledge.
- You had a plan and it was the wrong plan. Worth real time. This is where the difference between rating bands lives.
- You lost a fraction of a pawn in a quiet position. Ignore it. If a move is marked as an inaccuracy in a position that was never going to be decided by it, it is engine noise, not a lesson.
A useful filter: if the move did not change who was winning, it did not change the game. The labels are bands of exactly that measurement, so a blunder always means something changed hands and an inaccuracy often does not.
Do not learn the engine's line
An engine at depth 18 plays moves whose point arrives eight moves later. Memorising that line is worthless. What you want from it is the idea: it kept the bishop, it opened the file, it ignored the pawn because the attack was faster. If you cannot say in a sentence why the engine's move is better, the position is above where you are working right now, and there will be a cheaper mistake elsewhere in the same game.
Look at the opening once, at the end
Note the move where you left what you knew. Not to memorise the next ten moves of theory — just to know where your preparation ran out, so you can look at that one position properly later. One position per game is a real habit; twenty moves of theory per game is not.
Review the games you won
A win says nothing about whether you played well. Games you won are the ones where nobody punished the mistake, which means the mistake is still in your play and is entirely invisible to you. Accuracy on a game you won by forty moves of grinding is often lower than on one you lost.
Then do it again tomorrow
Ten minutes on one game beats an hour on six. The reason to have a review tool with no daily cap is not to review ten games at once — it is so that reviewing tomorrow's game costs nothing either.