For two years I thought I was stuck because I did not know enough. The next opening, the endgame I had not studied, the line I got caught in. So I watched more videos and read more theory, and my rating did not move. What finally moved it was embarrassingly simple. I started reviewing my games instead of immediately starting the next one.
I climbed from 600 to 1850 with no coach. Almost none of it came from new knowledge. It came from noticing that I kept losing the same few ways, and slowly losing them less often. Reviewing is how you notice. Here is the routine I settled on, and it takes about ten minutes a game.
Why most reviewing does nothing
Most players, when they review at all, click through forty moves, nod at the big red blunder they already knew about, and close the tab. That is not review, it is rewatching. You already knew you hung the rook on move 28. Confirming it teaches you nothing.
A review is worth the time only if it answers one question: where did the game actually turn, and would you find the right move next time? Everything below is built around that.
Step 1: write down what you were thinking, before the engine
Do this first, while the game is fresh and before any engine touches it. Scroll through and jot a line or two: where you felt confident, where you were guessing, the moment you thought you were winning, the moment it slipped. This is the part the engine cannot give you. The engine knows what was best. Only you know what you were actually thinking when you went wrong, and that is the thing you are trying to fix.
Step 2: find the turning points
Now bring in the engine, and do not read every move. Look for the moments the evaluation swung hard, a full point or more, in one move. Those swings are the whole game. A position can be equal for thirty moves and then decided in one. Find those two or three moments and ignore the rest.
The surprises matter more than the blunder you remember. A swing you do not recognise is more useful than the one you already knew about, because it is a hole you cannot see from the inside.
Step 3: ask "would I find it?" not "what is best?"
At each turning point, before you reveal the engine's move, try to find it yourself from that position. This is the most important step and the one almost everyone skips. There is a real difference between understanding the right move once it is shown to you and being able to find it at the board. Only the second one wins games.
If you can find it with no clock and a calm head, it was a one-off, move on. If you cannot find it even now, that is not a fluke. That is a gap in your pattern recognition, and it is a training target.
Step 4: look for the repeat, not the one-off
After a handful of reviews, the same themes start showing up. You hang pieces to back-rank ideas. You push too hard in equal positions. You fall apart once you are winning. The specific game does not matter much. The pattern across games is the whole point, and it is the thing worth drilling until it stops happening.
A single blunder you will never repeat is noise. The mistake you make in game after game is signal. Spend your time on the signal.
Keep it short, keep it regular
You do not need to review every game, and you should not try. Review your losses and the wins that were closer than the scoreline, a few games a week, ten minutes each. A short review you actually do beats a thorough one you dread and skip. Done often, this is the most reliable way I know to climb.
Where the tools come in
All of this is just a routine, and you can run it by hand. The reason I started building Lumichess was to take the manual parts off my plate. A Game Report does steps two through four for you: it runs Stockfish over the whole game, marks exactly where it turned, classifies every move, and lets you try the critical positions yourself before it shows you the answer. When you want the pattern across many games rather than one, a Profile Report runs the same analysis over roughly your last 40 and tells you which weakness keeps costing you. Review one game, find the pattern, drill it. That loop is the whole idea.
Frequently asked questions
How often should I review my chess games?
You do not need to review every game. Review your losses and the wins that felt closer than the result, a few games a week, for ten minutes each. Consistency matters more than volume.
Should I use an engine to review my games?
Yes, but second. First write down what you were thinking during the game, then let the engine show you where the evaluation actually turned. The engine tells you what happened; your notes tell you why.
What is the most important thing to look for?
The repeat, not the one-off. A blunder you will never make again is noise. The mistake you make in game after game is the one worth training.