How to analyze your chess games.

You learn from the review, not the game.

Playing tells you that something went wrong. Only reviewing tells you what — and whether it was a gap in your opening, a tactic you missed, or a plan you didn't have.

The trouble is that most reviews are either skipped or done badly: open the engine, watch the evaluation bar swing, nod at the best moves, close the tab. The engine was right; you learned nothing. The method below fixes that.

Step one: your own notes, before any engine.

Right after the game, while you still remember, go through it without help and mark the moments that mattered:

• Where you left your opening preparation, and whether you knew what to do next.
• The moves where you spent the most time — those are the positions you found hard.
• What you were thinking at each critical moment: the plan, the threat you feared, the line you calculated.
• Where you felt the game turn.

This is the step everybody skips, and it's the one that makes the rest useful. When the engine later disagrees with you, you'll know exactly which of your thoughts was wrong.

Step two: let the engine referee.

Now switch it on. Step through the game and compare each verdict with your notes. Below is a short game with Stockfish's real evaluations — use the arrows, click a move, or click the graph. Where a move goes wrong, the green arrow shows what the engine preferred.

What the names mean

Engines measure each move by how much it changed your chances of winning compared with the best move. A small drop is an inaccuracy, a larger one a mistake, and a drop that usually changes the result is a blunder. Look at the graph. Black's 3…Nd4 hands White an edge, White's greedy 4.Nxe5 gives it straight back, and 5.Nxf7?? decides the game. Two of those three cliffs are captures that "won" material.

Step three: find the turning point.

Don't try to fix every inaccuracy. Find the one or two moments where the evaluation swung hardest — the cliffs in the graph — and give those all your attention. For each one, ask:

Step four: turn the lesson into training.

A review that ends with "I should have played Nd5" is half finished. Every turning point should leave something behind that you will practise.

Look for patterns across games

One game is an anecdote; twenty are data. If the same kind of mistake shows up again and again — missed knight forks, a weak back rank, trouble in one opening — that's where your next month of training should go. Reviewing many games at once is exactly what Blind Spots does: it reads your recent games and groups what you missed by pattern.

Questions, answered.