Everyone says "analyze your games." Almost nobody says how. So most improving players do one of two things: skip it entirely and queue the next game, or open an engine, watch it spit out lines like 18.Bxf7+!! Rxf7 19.Qd5, understand none of it, and close the tab feeling vaguely guilty. There's a better way, and it takes about ten minutes per game.
Step 1: Guess before you look (2 minutes)
Before any engine sees the game, ask yourself: where do I think it went wrong? Write down one move number, even just in your head. This tiny step is where most of the learning lives, because it tests whether your sense of the game matches reality. Being wrong about where you lost is the most useful thing a review can teach you.
Step 2: Find the two or three real moments (5 minutes)
Now run the analysis and look at the evaluation graph, not the move-by-move labels. Find the biggest swings, the cliffs where the line jumps. A typical game has two or three that matter; the rest is noise. For each cliff, ask three questions. What did I think I was doing? What did the position actually demand? And what would have made me see it at the board?
That third question is the one that separates useful review from ritual. "I should have seen the fork" isn't an answer. "I never checked what his knight could jump to after I moved my rook" is an answer, because it's a checkable habit you can apply next game.
Step 3: Name the mistake (2 minutes)
Every mistake worth remembering has a name: hung a piece, missed a fork, weakened the back rank, pushed a pawn in front of my own king. Naming it matters because your brain files named patterns and retrieves them under pressure; it does not retrieve vague guilt. If you can't name it, it wasn't really a lesson yet.
The part everyone skips: memory across games
Here's the uncomfortable truth about single-game review: it evaporates. You review Tuesday's game, learn Tuesday's lesson, and by Friday it's gone, because one exposure to a pattern isn't how memory works. The players who actually convert review into rating keep some record of their named mistakes and re-check it, so that the third knight fork gets recognized as "this again," not experienced as brand new.
That's the loop Rookie automates: every reviewed game's mistakes get named, tagged, and kept in a deck that spots the repeats for you, then turns them into puzzles from your own positions. But tool or no tool, the method stands: guess first, find the cliffs, name the mistake, and keep the names somewhere they can confront you later.