Review a game by finding the first decision you did not understand, then distinguish it from the first serious error. Leaving memorised theory is not automatically a mistake. You may have chosen a perfectly playable move and gone wrong much later. Use your own notes and the board before consulting an engine so that you can identify what you actually thought during play.
Classify the difficulty. A recall error means you forgot a prepared continuation. A calculation error means you missed a forcing response. A planning error means your moves did not fit the structure. These problems need different work: spaced recall for the first, tactical exercises for the second, and model games or structural study for the third.
In the Two Knights position shown, Black's ...Nxd5 allows Nxf7 and a possible attack on the exposed king. If you missed that idea, study the forcing line and a reliable alternative such as ...Na5 at the previous decision. Simply replaying the first four moves would not address the tactical misunderstanding. Record the precise trigger that should make you pause next time.
Write one useful correction after each game: the position, your candidate, the better idea and the reason. Check the concrete line with a trusted analysis source when necessary. Add the relevant branch to practice and review it later. Track repeated patterns across several games; they tell you far more about your next study priority than a single opening win rate.
Try this at the board
Take one mistake from your latest game and write a one-sentence explanation without using an engine evaluation number.