Review your own game before looking at the evaluation

Start with the decisions you remember

Replay the game on a board without analysis running beside it. Mark the moments when you spent time, changed your mind or made a move mainly to escape an uncomfortable position. A move that was objectively sound can still reveal a poor thinking habit. Conversely, a losing move may come from a sensible plan with one concrete detail overlooked.

Write a sentence beside each marked position. I wanted to exchange the active bishop is useful. This move was bad is only a verdict. Include what you expected the opponent to do, especially if the actual reply surprised you. That difference between expectation and reply often identifies the question you forgot to ask at the board.

Reconstruct the alternatives fairly

Choose two or three candidate moves that you could reasonably have considered during the game. Do not fill the page with every legal move. For each candidate, give the opponent a purposeful reply before extending the line. If your attractive attack requires the opponent to move a piece away for no reason, the variation has already stopped being a useful test.

Keep the depth modest and explain the resulting position. Who has the more active pieces? Is an exchange likely to help one side? Which pawn break changes the character of the game? These descriptions make it possible to compare two short lines without pretending you can calculate every quiet branch to a final numerical answer.

• Record two or three plausible candidates.
• Look for the opponent's useful reply.
• Describe the position at the end of the line.

Use the engine to locate the broken assumption

Only after the notes exist should you introduce an engine. Begin at the marked positions instead of watching the evaluation flicker through the entire game. When the machine dislikes a move, inspect the response it recommends. The important discovery might be an undefended piece, a forcing intermediate move or an endgame that your written description treated too casually.

Do not copy a long principal variation into the notebook and call the position understood. Replay the first few moves and ask what makes them work. Try the natural alternative you originally expected. If you cannot explain why that alternative fails, keep the question open. A short unresolved question is more honest and more useful than a confident page of borrowed notation.

End with one change to your next game

Group the mistakes by cause. An opening memory lapse, a calculation error and a rushed move in time trouble require different work. Counting them all as blunders hides the distinction. Choose the recurring cause that you can address with a concrete habit, such as checking loose pieces before committing to a pawn break or reserving time for the first unfamiliar position.

Finish the review with one position to revisit and one sentence to remember. Keep the original notes, including the mistaken explanation, so that the next review shows whether the habit changed. An evening spent understanding three decisions is enough. The purpose of the record is to improve the next decision, not to produce an impressive quantity of annotated moves.