How to Review Your Chess Games: A 20-Minute Routine
Most players review a game the same way: click "analyze", scroll through the red marks, think "I see", and move on to the next game. Nothing changes afterwards, because nothing was decided. This routine takes about twenty minutes and ends with a written lesson you can actually act on.
From the team. When one of us went through their own games in bulk, the mistakes were not random: they clustered when the clock was low, and the same few openings kept turning up. That is why step 4 below asks you to name the cause of each critical moment, not just the move.
Why most reviews change nothing
An engine is very good at telling you what the best move was. It is silent on the questions that matter for improvement: why you didn't play it, and whether you are likely to make the same kind of mistake again. If a review only answers the first question, you learn about one position in one game and forget it within a week.
The routine below keeps the engine, but puts it in the middle of the process instead of at the start. It works for a ten-minute blitz game or a long classical one, and it is short enough to do after every game you care about.
Step 1: Pick the game and set a timer
You don't need to review every game. Choose one:
- A loss you didn't understand. If you already know why you lost (you flagged, you hung a queen to a one-move threat), a long review adds little.
- A win that felt lucky. Wins hide as many problems as losses, and they are much less painful to look at honestly.
- A game against a slightly stronger opponent, where the distance between you and them is most informative.
Set a timer for twenty minutes. A time limit forces you to look for the two or three moments that decided the game rather than dissecting every move.
Step 2: Replay it before the engine sees it (3 minutes)
Play the moves through on a board with the engine off. Every time you remember a moment of doubt, write down the move number and what you were thinking. Also note where you felt the game turn, even if you can't say why.
This is the part that no software can do for you. The engine knows the position; only you know what you were thinking during it. The gap between the two is where the useful lessons live. A move you played confidently and that turns out to be a blunder points to a wrong belief. A move you played nervously that turns out to be fine points to a lack of confidence. Those need different fixes.
Step 3: Find the critical moments (5 minutes)
Now turn the engine on and look at the evaluation graph, not the move list. Games are usually decided by two or three large swings; the rest of the moves are noise around them. Find those swings and ignore anything smaller. In RookSac's terms that means concentrating on moves that lose more than about 10 win-percentage points (the Mistake and Blunder labels, plus Miss for a spoiled win) and skipping the smaller inaccuracies for now. See move classifications for the thresholds.
Compare the swings with your own notes from step 2. There are usually three kinds of match:
- You doubted it and the engine agrees. Your instincts are working. Good.
- You were sure and the engine disagrees. This is the most valuable kind, because it shows a belief that is wrong.
- You didn't notice anything and the engine shows a swing. You missed something outright: a tactic, a threat, a change in the position.
Step 4: Name the cause of each moment (7 minutes)
For each critical moment, ask what you saw, what you missed, and then put the mistake in one of a small number of categories. The categories matter because they point to different training:
| Cause | What it looks like | What to train |
|---|---|---|
| Tactical oversight | A piece left en prise, a fork or pin you didn't see. | A pre-move safety check and tactics puzzles. |
| Opening knowledge | You left theory early and fell for a known trap or a bad structure. | Learn the line and why the good move is good. |
| Positional misjudgment | No tactic, but you traded the wrong pieces or gave up a key square. | Model games and plans for the structure. |
| Endgame technique | A won or drawn ending that you didn't convert or hold. | Basic endgames, practised against an engine. |
| Time trouble | The error happened with under a minute or so left. | A time budget for each phase. |
| Emotional | You played fast or recklessly after a bad move. | A reset habit: stand up, breathe, then reassess. |
Here is a concrete example. In the Italian Game, after 1.e4 e5 2.Nf3 Nc6 3.Bc4 Nf6 4.Ng5 d5 5.exd5, the natural-looking recapture 5...Nxd5 is an inaccuracy. It walks into 6.Nxf7, the start of the Fried Liver Attack, where White gets a dangerous initiative for the sacrificed piece. The usual move is 5...Na5, kicking the bishop away.
The right label for this mistake is not "tactic". A player who classified it that way would spend a week on fork puzzles. The actual fix is ten minutes of learning the Two Knights Defence and the 5...Na5 line. Naming the cause is what stops you training the wrong thing.
Step 5: Write one sentence (2 minutes)
For each critical moment, write a single instruction to your future self in the form "When X, check Y." For instance:
- "When I move a piece that is defending something, check what it stops defending."
- "When I'm a pawn up, check whether my opponent's most active piece can be traded off."
- "After 1.e4 e5 2.Nf3 Nc6 3.Bc4 Nf6 4.Ng5, play ...d5 and then ...Na5, not ...Nxd5."
One or two sentences are enough. If you can't write one, you haven't found the cause yet. Go back to step 4. Keep them in a plain text file and reread the last ten before your next session.
Step 6: Keep a tally
After the review, add one line to a running list with the date, the opponent's rating, and the cause categories you found. After ten or so games, count them. One category will usually dominate, and that is what you should study next. The tally replaces the vague sense of "I keep blundering" with a specific target, such as "seven of my last ten critical moments were tactical oversights in the middlegame".
This is the same idea that reviewing many games at once automates: single games are anecdotes, and patterns need many games.
Common traps
- Only checking the engine's top line. If you can't explain in a sentence why the engine's move is better, you haven't learned anything from it yet. Play the line out and look for the reason.
- Learning engine-only moves. Some best moves are ones no human at your level would find. If a move needs a ten-move calculation to justify, note it and move on.
- Reviewing only losses. You'll overlearn from disasters and never notice the inaccuracies that shave points off every win.
- Over-weighting the opening. Below about 1500, a large share of games are decided in the middlegame by a single missed tactic, not by twelve moves of theory. Look at where the big swings actually happened, and don't assume it was the opening.
- Trusting a won evaluation. When you are already clearly winning, several moves are "fine" and the engine's number barely moves. That is why RookSac measures win-percentage change rather than raw centipawns; see how to read engine evaluations.
A checklist you can copy
- Choose one game. Set a 20-minute timer.
- Replay it without the engine. Note the moments of doubt and where the game turned.
- Turn the engine on. Find the two or three biggest swings.
- For each: what did I see, what did I miss, which cause category is it?
- Write one "When X, check Y" sentence per moment.
- Add the categories to your tally.
Do this for a month and you will have a written record of your own recurring mistakes, which is more useful than any opening book or list of tips.