Write Down Your Own Thinking Before You Touch an Engine

Write down your own thinking before you touch an engine

Here is the direct answer. Go through your game move by move. Write down what you were actually thinking. What did you look at? What did you skip? Where did you feel unsure? Do all of that first. Only then turn on the engine. Check your notes against it second.

GM Illingworth lays out a simple three-step process for this. Step one: find the mistakes. Find the missed improvements too. Look at both sides, not just yourself. Step two: figure out exactly what you needed to see in that position. What would have shown you the better move? Step three: use that answer to change how you train. Make the same kind of spot go better next time.

The order matters. If you fire up the engine first, you see the best move right away. But you have not yet written down what you actually saw at the board. That erases the evidence. You cannot fix what stopped you from finding a move if you never wrote down what you were looking at when you missed it.

So before any engine line, sit with the game on your own. Say out loud, or type, what worried you on move 12. Or why you skipped a capture on move 20. That record is what step two and step three actually run on.

Seeing the engine's best move doesn't teach you anything by itself

GM Illingworth makes a point that matters here. Knowing the engine's top move is not enough. You need to know why you did not find it.

Most of the time, the reason is simple. Maybe you never even looked at that move. It never became a candidate move, meaning one of the options you seriously weighed before playing. Or maybe you looked at it and misjudged the position. Or you missed something your opponent could do in reply. The engine will show you the right answer. It will not show you which of these happened in your head.

That is why self-annotation has to come first. If you check the engine right away, it hands you the answer before you have rebuilt your own thinking. Ten minutes later, you will remember the engine's line. You will not remember your own reasoning. The moment is gone.

So before you touch Stockfish, sit with the position from memory. Stockfish is the free chess engine most sites use. Ask yourself: what moves did I actually look at? Say you played a rook move and missed a knight jumping into the center. Write that down now, while you can still recall why the knight never crossed your mind. Maybe you were focused on the kingside. Maybe you stopped scanning the rest of the board. Maybe you assumed a square was safe without checking it.

That memory is the only copy you will ever get of what you actually saw at the board. Once the engine speaks, it is gone.

Your own analysis will still miss real tactics — that's normal, not a reason to skip it

Here's the honest part. Self-annotation will still miss things. That's writing down your own thinking, move by move. Missing things is not a reason to skip it. It's just what step one is for.

A Lichess forum thread talked about analyzing without an engine. One player described doing exactly this. They reviewed their own games with no engine at all. Then they showed that analysis to a player rated 2100. The 2100 player found tactics the original player had missed. Missed twice, in fact. Once during the actual game. And again during their own careful post-game review. Two chances to see it. It stayed hidden both times.

It gets more specific than that. Another player told a story on a Chess.com forum thread. They annotated a slower game by hand, played at 30 minutes plus 20 seconds a move. No engine. They labeled one of their opponent's moves a blunder. A blunder is a move that loses material or the game right away. A 2100-rated reviewer looked at the same position. That reviewer said the move was probably the best one on the board. The label was backward.

That same Lichess thread pointed out something bigger. Some of the most famous chess analysis ever published has later been shown wrong. Strong players wrote it about their own games. Engines checked it later and found the holes. If that happens to strong players writing about their own games, it will happen to you too.

None of this means self-annotation is a waste of time. It means it's a first draft of the truth, not the final one. You write down what you saw and what you were thinking. Then you check it. The checking step is what catches the tactic you and everyone else missed at the board.

What the engine disagrees with tells you exactly what to train next

Now open the engine. Line it up against your own notes, move by move. Every place you disagree with the engine is a lesson. GM Illingworth's fix map tells you which lesson it is.

Did you never even consider the engine's best move? Not as a candidate move, meaning one of the options you weighed before playing. That's a calculation gap. The fix is practicing how you generate candidate moves. Then the right one shows up on your list next time.

Was the loss a silly blunder? A move that hands over material or the game right away, one you'd catch instantly if someone else played it. That's tactical vision. The fix is more tactics training. Not more opening study.

Did you feel uneasy in a position the engine says was actually fine for you? That's a structure you don't understand yet. The fix is studying that specific pawn structure or piece setup. Not grinding more games blind.

Did a trade go wrong, leaving you worse than you expected? That's exchange evaluation. The fix is practicing whether a trade helps or hurts you before you make it.

This is where BlackBox's real numbers connect. Missing a fork you could have played is called Fork Miss (O) in BlackBox's data. It hits 65% of players rated 1200–1399. About 27 times each, at severity 6.4. That's not bad luck. It's a candidate-move gap. The fork was on the board. It never made your list of options. The same goes for Rook Activation, Missed 7th Rank (O), at 67% of players, severity 5.5. The winning idea existed. You just didn't generate it as a candidate. The engine disagreement doesn't just tell you that you were wrong. It tells you which training bucket you were wrong in.

This routine works game by game — finding the pattern needs a batch

Here's the catch. Self-annotate, then verify with an engine. That's the right routine for one game. It won't show you what keeps happening across ten games. Or a hundred. For that, you need to put many games side by side. Not study them one at a time.

Think about it this way. A fork is one move that attacks two of your pieces at once, so you can only save one. Say a fork costs you a game on a Tuesday. Your single-game routine catches it. But it can't tell you that the same fork cost you three more games last week. Only a batch view across games can show you that.

This is exactly what BlackBox Chess does. It's a tool built for chess players under 2000 Elo who want to fix what's actually costing them rating. The free check runs Stockfish over your last 10 games. It names your #1 recurring mistake. No account, no card. The $10 Full Report goes deeper. It covers 100 games. It ranks every mistake pattern by severity. It includes three real game examples and a fix course built around your top pattern.

That's not a hypothetical benefit. In BlackBox's real data, one 1200-rated player had Fork Exposed (D). That means they kept getting forked on defense. It happened 11 separate times, at severity 8. First caught in game 8, at move 13. A single-game review would have caught each fork alone. Only the batch view showed it was the same leak, over and over. If that sounds like your last ten losses, the free check takes a few minutes. It tells you where to look first.