How to Analyze Your Chess Games: A Repeatable System

Most players know they should analyze their games. Very few actually do it, and of those who do, most open the engine, watch the evaluation bar jump around for two minutes, mutter "yeah, that knight move was bad," and close the tab. That's not analysis. That's watching a replay of your own crash without ever asking why you steered into the wall.

The good news is that useful analysis doesn't require a coach, a database subscription, or hours per game. It requires a system: the same short sequence of steps, applied to every serious game you play. Here's one that works, and why each step earns its place.

Step 1: Pick the right games

Not every game deserves a full review. A 1-minute bullet game where you both flagged in a rook endgame tells you almost nothing about your chess. Prioritize:

• Losses and draws in longer time controls (10 minutes and up). These are the games where you had time to think, which means your mistakes reflect your actual understanding, not your mouse speed.
• Wins that felt uncomfortable. If you won because your opponent hung a queen on move 30, but you were worse for the first 25 moves, the win is hiding a loss's worth of lessons.
• Games where you remember a specific moment of confusion. "I had no idea what to do after he played h5" is a flag worth planting.

One well-analyzed game per day beats ten skimmed ones. If you play a lot of blitz, pick your one most instructive game and let the rest go.

Step 2: Annotate before you open the engine

This is the step almost everyone skips, and it's the one that matters most. Before any engine sees the game, play through it yourself and write down three things, even if it's just a phrase per moment:

1. Where you left your preparation. The first move where you were thinking for yourself instead of playing something you knew.
2. What you were thinking at the key decisions. Not what you now believe was right, but what you actually thought at the board. "I played Nd5 because I was worried about his bishop coming to b4" is gold, even if Nd5 was a blunder.
3. Where you felt the game turned. Your gut sense of the critical moment, before the eval graph tells you the official answer.

Why bother? Because the engine can only tell you what was wrong. Only your own notes can tell you why you played it anyway, and the gap between those two things is exactly what you need to train. If the engine says 14...Nxe4 loses to 15.Bxf7+ and your note says "I saw Bxf7+ but thought I could just take back," the lesson isn't "Nxe4 was bad." It's "I stopped calculating one move too early." That's a fixable habit. "Nxe4 was bad" is just trivia.

All of this takes maybe ten minutes, and nothing else in the process pays off as much.

Step 3: Run the engine, but read it correctly

Now let the engine analyze the game. Any strong engine will do; Stockfish is free and runs in your browser on most chess sites. What matters is how you read the output.

Look at eval swings, not eval levels. An evaluation of +1.5 means little on its own. A jump from +0.3 to −2.1 in one move means everything. That's a critical moment by definition. Scan the graph for the three or four biggest swings and mark those positions. Those are your critical moments, whether or not they match the ones you guessed in Step 2. (When they don't match, that mismatch is itself a finding: the game turned somewhere you didn't notice, which usually means a threat you never saw.)

Ignore small wobbles. The difference between the engine's first choice at +0.4 and your move at +0.1 is not a mistake worth studying at club level. Chasing every third-decimal improvement turns analysis into noise. You're looking for the moves that changed the result, not the moves that offended the silicon.

Ask "what was the threat?" before "what was the best move?" At each critical moment, before you look at the engine's suggestion, look at the position and find what your opponent was actually threatening or what resource you missed. Suppose you castled short in an Italian, played a natural-looking 12...b5, and the eval dropped two points. Don't just note that the engine wanted 12...h6. Find the punishment: maybe b5 loosened c6 and allowed Nd5, hitting your bishop on e7 and the c7-square at once. The refutation teaches you a pattern; the engine's preference teaches you nothing you'll remember on Tuesday.

Step 4: Name the mistake

For each critical moment, write one line that classifies what actually went wrong. Most club-level mistakes fall into a handful of buckets:

• Missed opponent's threat: you calculated your idea and never asked what their last move did.
• Stopped calculating too early: you saw the line but cut it off a move before the tactic landed.
• Wrong plan: no single move was terrible, but you spent five moves attacking on the kingside while the game was being decided on the c-file.
• Known position, forgotten idea: you'd seen this structure before and didn't recall how to handle it (the classic example: facing the London and never getting ...c5 in).
• Time trouble: the move was bad because you had 40 seconds, which is really a clock-management mistake from move 20, not a chess mistake at move 38.

The bucket matters more than the move. Three games in a row with "missed opponent's threat" tells you exactly what to train: before every move, one forced habit. What did their last move change? No opening course will fix that; awareness will.

Step 5: Turn it into one lesson you'll keep

End every analysis session by writing a single takeaway in a running note: a physical notebook, a document, wherever you'll actually look again. One sentence, concrete, in your own words:

"In the Advance Caro-Kann, stop auto-playing ...c5 before deciding whether the knight belongs on d7 or a6."

"When I'm winning, I start playing fast. Slow down after winning material, that's when I hang things back."

Not "calculate better." Not "study endgames." A real lesson names a situation you'll recognize and the behavior to change when it appears. Reread the list before you play. After twenty games you'll have a page that describes your chess more accurately than any rating graph, and you'll start catching your signature mistakes in real time, which is the whole point.

The system, in one glance

1. Pick one instructive game (losses first, longer time controls).
2. Annotate it yourself before the engine sees it: your thoughts, your critical moment.
3. Engine pass: find the big eval swings, ignore the wobbles, find the refutation, not just the better move.
4. Name each mistake's category, not just the move.
5. Write one concrete lesson in a running list. Reread it before you play.

Twenty to thirty minutes per game. The players who do this consistently improve faster than the players who play three more blitz games in the same half hour. A mistake you understand stops recurring, while a mistake you merely regret shows up again next week wearing a different opening.

Where a coach fits in

The hardest part of this system is Step 3 and 4 when you genuinely can't see why the engine's move is better: the eval says −2, you stare at the position, and the reason won't come. That's the moment a coach helps: someone who can say "the problem isn't the knight, it's that your dark-squared bishop died on b2 three moves ago." If you don't have a human coach on call, Cody does this from your browser: a free account lets you import a finished game from a public Chess.com or Lichess link or a PGN, including over-the-board games, and builds a move-by-move review you can question from the board. Ask why a move failed and it explains the plans and threats in plain language, with arrows on the board, by text or voice. Text coaching is free, so you can keep asking why until it clicks. This review loop is also the center of a broader chess improvement plan. See how Cody's free chess game review works.