To analyze a chess game, first play through it once without the engine and note where you were unsure. Then import it, run the engine at a consistent depth, and review only the handful of moves that changed the evaluation. Finish by writing one sentence about the lesson and one thing to practise.
Most players analyse the wrong way: they load the engine, watch it disagree with them, and move on. The engine tells you where you went wrong, but it cannot tell you why you did not see it. That part is your job, and it is where the improvement actually happens.
Step 1: play through it yourself first
Before any engine touches the game, replay your own moves. This is the single most skipped and most valuable step. While you go, mark the points where you remember being unsure, or where you had to choose between two moves.
Step 2: choose one depth and keep it
Decide on a search depth before you start and use it for the whole game. A review at mid-teens depth finds nearly every practical error and finishes quickly. Raising the depth only for the positions that mattered most is a good compromise, and it keeps the other ninety percent of the game fast.
Consistency is the point. Grading half a game at depth 10 and half at depth 30 gives you a move list whose labels are not comparable with each other.
Step 3: find the turning points, not every mistake
Scan the evaluation graph for the places where the bar crossed from one side to the other or changed sharply. Those are three or four moments in a typical game. Everything else is noise you can safely skip.
- A move that turned a comfortable position into a difficult one.
- A moment where you had a strong continuation and missed it.
- A position where you spent most of your time, regardless of the result of the move.
- Any point where you realised your plan had run out.
Step 4: ask the right question at each one
Do not ask what the engine would have played. Ask why you did not play it. The answers fall into a small number of categories, and each one points to a different kind of practice.
| Answer you give | What it means | What to work on |
|---|---|---|
| I did not see it at all | A blindness or scanning problem | Slow down and list candidate moves |
| I saw it and rejected it | A judgement error | Calculate the line you feared to the end |
| I had no plan | A positional gap | Study plans typical of the structure |
| I was short of time | A time management issue | Practise the opening you played |
| I knew it was bad but had nothing else | A hole in your knowledge | Study that specific pattern |
Step 5: turn it into something to practise
An analysis you do not convert into practice is entertainment. From the turning points, pull out concrete positions: the one where you missed the tactic, the one where you chose the wrong plan. Those positions are the raw material for your own puzzle set.
Saving positions to replay a week later is far more effective than reading an engine line once. Recognition is built by repetition, and you cannot repeat what you did not save.
A note on accuracy percentages
Accuracy summaries are computed from the same evaluations as the individual grades, so they carry the same caveat: they describe a specific analysis at a specific depth. A high accuracy in a quiet position where every move was easy is not the same achievement as a lower accuracy in a sharp one, and a single blunder can swing the percentage more than several ordinary inaccuracies.