How to analyse your own games with a chess engine

A typical review looks like this: load the game, run the analysis, scroll through the moves, see the red marks, nod, close the tab. A week later the same mistake comes back.

The reason is that recognising someone else's answer does not build a skill. The reverse order works:

1. 1. Go through the game yourself, with no engineMark the places where you had to think, where you were afraid, and where you believe you went wrong. Write your guesses down.
2. 2. Only now turn the engine onAnd compare: did your suspicions match its verdict? The most valuable finds are the errors you did not notice at all.
3. 3. Work out why the move was badDo not memorise «a knight was needed here» - understand the mechanism: what exactly did I fail to see. Without this step the review is worthless.

The difference between the two approaches is enormous: the first takes three minutes and gives nothing, the second takes twenty and is worth several games played.

The evaluation is given in pawns: +1.0 means White is about a pawn better, −2.5 means Black is about two and a half pawns better. Mate in N is shown separately.

The key caveat: an engine evaluation is an evaluation for an engine. A position at +0.8 where you must walk twenty moves along a single path is objectively worse for an amateur than an equal position with a clear plan. Do not transfer the machine's verdict onto yourself automatically.

Analysis services usually label moves by how much evaluation they lost. The names differ, the meaning is the same:

• Inaccuracy. The evaluation slipped a little. For an amateur this rarely deserves attention at all.
• Mistake. A noticeable loss: a win let slip or a position spoiled.
• Blunder. The game-turning kind: a piece hung or a tactic missed.

Practical advice: review only blunders and mistakes. Chasing inaccuracies at amateur level is a waste of time - they do not change results and they dilute attention.

Look separately at your opponent's moves: their mistakes that you missed are your material too. The moment they left something hanging and you did not notice deserves as much attention as your own blunder.

• Why you played that move. It sees the move but not that you missed a pin or were simply rushing.
• What was practically hard. A move the engine calls the only one may be impossible for a human to find - yet the review marks it as your error.
• Where your time went. Twenty minutes on an obvious opening move is a serious problem the engine will not notice at all.
• A plan. It gives the best move, not the idea. Working out what is actually going on remains your job.
• The psychology of the game. The moment you fell apart after losing a pawn does not appear in any evaluation.

So the best review is a combined one: your own thoughts first, then the engine, and ideally a conversation with your opponent or a coach afterwards.

The most underrated tool there is. Reviewing one game gives you a fact; reviewing ten in a row starts to show a pattern - and the pattern is what deserves your time.

Keep a simple file and add one line after every review:

After a month the list itself tells you what to study: the same line appearing four times is your topic for the near future. It beats any study plan, because it is about you rather than about an average player.

• Do not review everything. All the games of an event is too many. Take two or three: the most painful loss and the game you understood least.
• Losses are worth more than wins. Though wins are worth checking too - they often turn out to have been accidental.
• Twenty minutes per game. Longer and attention drops, and the review turns back into scrolling.
• Write your games down at events. Without a scoresheet there is nothing to review.
• Free tools are fine. Any modern engine is orders of magnitude stronger than you; a paid subscription adds nothing essential at this level.
• Go back to old reviews. Six months later the same games read differently - the best measure of your own progress.