How to Analyze Your Chess Games

To analyze your chess games properly, review each game in four steps: first without an engine, writing down what you thought and where you think you went wrong; then answer three diagnostic questions about where the game swung and why; then run an engine pass to check your conclusions; and finally file the lesson in a journal you review monthly. As a grandmaster and coach, I can tell you that players who do this with two games a week improve faster than players who do ten hours of unfocused study — because chess analysis of your own games is the only training material written specifically about you.

Why Analyzing Your Own Games Is the Highest-Leverage Habit

Every serious chess school agrees on one thing, and it traces back to Mikhail Botvinnik. The sixth world champion built his famous school — the one that trained Karpov, Kasparov, and Kramnik — around a single demanding ritual: students analyzed their own games in depth and presented that analysis for criticism. Botvinnik believed, and modern coaching confirms, that a player's own games are a mirror no book can replace.

Here is the logic. A book on tactics teaches the tactics its author chose. A video on pawn structures teaches the structures its presenter likes. But your own games contain your exact mistakes, in the exact positions you actually reach, made under the exact time pressure you actually feel. No published material is that personalized. When I train students, the first thing I ask for is not their rating or their opening repertoire — it is five of their recent losses. Those five games tell me more than any questionnaire.

The habit also compounds. Fix a recurring leak — say, you always drift in symmetrical endgames — and every future game benefits. That is why game analysis sits at the top of the priority list in my guide on how to get better at chess: it is the step that makes all the other training targeted instead of random.

Step 1: Analyze Without the Engine First

This is the step almost everyone skips, and it is the one that produces the improvement. Before you let any engine touch the game, go through it yourself and reconstruct your thinking. Write or say out loud: what was my plan here? What did I consider? Why did I reject the alternatives? Where did I first feel the game slipping?

Then try to improve on your own play. At every moment you remember being unsure, generate candidate moves fresh, as if it were a training exercise, and work out which one you now believe is best — with reasons, not feelings.

Why does this matter so much? Because analysis without the engine is a workout for the same muscle you use during the game: unassisted judgment. When you switch the engine on first, you stop evaluating and start guessing what the machine will say. You become a spectator of your own game. I have watched students click through engine lines for an hour and retain nothing, because they never formed an opinion the engine could correct. The correction is where learning lives — and you cannot be corrected if you never committed to a view.

Practical format: fifteen to thirty minutes per game, notes in a file or notebook, three to five moments flagged as "critical — check later." That flag list is the input for Step 3.

Step 2: Ask Three Questions of Every Game

Unstructured analysis drifts into replaying your favorite moments. Structure it with three questions, answered in writing.

1. Where did the evaluation swing? Not "where did I blunder a piece" — where did the game actually change character? Often the real turning point is five moves before the visible mistake, when you chose a passive plan or released the tension for no reason.

2. What did I miss or misjudge — and why? The "why" is the whole point. A missed tactic can come from three different diseases with three different cures: a calculation error (you saw the line but evaluated its end wrongly), a pattern gap (you never considered the motif at all), or time pressure (you saw nothing because you had ninety seconds). Writing "missed Nxf7" teaches you nothing. Writing "never considered the sacrifice because I don't know Greek-gift patterns" gives you a to-do item.

3. What ONE pattern do I extract? One, not five. A single transferable sentence: "In opposite-side castling, I count attackers before grabbing pawns." "When my opponent has the bishop pair, I close the position before simplifying." If you cannot compress the game into one lesson, you analyzed the moves but not yourself.

Step 3: Now the Engine Pass

With your own conclusions written down, the engine becomes what it should be: a tactically infallible sparring partner that checks your work. Load the game on our free analysis board and go through it once more, comparing the engine's verdicts against your Step 1 notes.

You are looking for two kinds of discovery. First, tactical truths you both missed during the game — the shot that was available on move 23 for either side. Second, and more valuable, disagreements between your Step 1 opinion and the engine: every position where you wrote "this is fine for me" and the engine says −1.8 is a hole in your evaluation skill, precisely located.

A rule I give every student: treat any evaluation swing bigger than about 1.5 pawns as a study position. Save the position, set it up later cold, and try to find the right idea again from scratch. Small wobbles of ±0.3 are noise at club level; the 1.5+ swings are where games are actually won and lost, and they deserve real attention.

Keep the engine pass shorter than the human pass — twenty minutes is plenty. If the machine time exceeds the thinking time, the ratio is backwards.

Step 4: File the Lesson in a Leak Journal

Analysis that is not recorded evaporates. Keep a leak journal: one line per game with the date, the one extracted pattern from Step 2, and the cause category (calculation, pattern gap, evaluation, clock, opening).

Then review it monthly, because the monthly review is where the magic happens. Individual games look like random accidents; twenty journal lines reveal themes. Is it always the same pawn structure where you drift? Do half your losses list time trouble as the cause — meaning your real problem is decision speed, not chess knowledge? Do your collapses cluster right after the opening ends, when the memorized moves run out? Each of these themes points to a completely different training plan, and you cannot see any of them one game at a time.

In my own career this was humbling: my journal once showed that a large share of my worst decisions came in positions where I stood clearly better — I relaxed. No book would ever have told me that.

Which Games to Analyze First

Your losses, before your wins. Losses contain the leaks; wins mostly contain your opponent's leaks. Analyzing wins feels better, which is exactly why it teaches less — you replay your good moves and skim past the moments your opponent let you off the hook.

Long games over fast ones. A classical or rapid game records your genuine thinking; you had time to form plans, so the mistakes reflect your real understanding. A blitz game mostly records your reflexes — worth a two-minute skim for recurring opening problems and hung pieces, but not a deep session.

Painful games especially. The loss you do not want to look at — the winning position thrown away, the opening disaster against a lower-rated player — is almost always the most instructive game in your collection. The pain is information: it marks the spot where your self-image and your actual play disagree. Analyze that one first, while it still stings a little.

Fixing Your Openings From Your Own Games

Most players repair their repertoire backwards: they study more theory in the abstract, then keep losing in the same lines. Reverse it. Your analyzed games show you exactly where your repertoire actually broke — the move where you were out of book, uncomfortable, and worse by move 14.

For each opening problem the journal surfaces, do three things. Identify the precise branching point where your knowledge ended. Check whether the position arose directly or via transposition — club players are surprised how often their "anti-Sicilian problem" is really a move-order problem. Then drill the fixed line until it is automatic; our opening trainer exists for exactly this, letting you rehearse the correct moves against realistic replies instead of just reading them.

This is repertoire maintenance the way professionals do it: driven by your games, not by fashion.

Reading the Eval Graph: Spikes vs Squeezes

The evaluation graph of an analyzed game has a shape, and the shape is a diagnosis.

A blunder spike — a flat line that suddenly cliffs — means the game was decided by a single visible blunder. The cure is tactical: safety checks before every move, pattern training, clock management so the critical moment does not arrive with seconds left.

A slow squeeze — a line that drifts from equal to lost over fifteen moves with no single dramatic drop — is a different disease entirely. Nothing "happened," yet you lost. That graph says your opponent understood the position better than you did: you misplaced pieces, chose the wrong plan, made small concessions. The cure is strategic, not tactical — better plans, better piece placement, the ideas I cover in my guide to chess strategy.

Players who only hunt for the big red blunder in their games systematically misdiagnose squeeze losses as "I played fine and then blundered." No — the blunder at move 38 happened because you were already suffering for twenty moves. The graph knows the difference; learn to read it.

Common Analysis Mistakes

Engine worship. Treating every engine preference as a lesson. If the engine says your move was +0.3 and its move was +0.6, at club level that difference is meaningless — both moves are fine, and agonizing over it is wasted study time. Reserve your attention for the big swings.

Skipping the why. Cataloguing mistakes without diagnosing causes. "I blundered on move 25" is a fact; "I blundered because I stopped checking my opponent's threats once I was winning" is a lesson. Only the second changes your next game.

Only analyzing wins. Comfortable and nearly useless, as above.

Engine-first analysis. Switching the machine on immediately, which converts a thinking workout into a guessing game and quietly deletes the main benefit of the whole exercise.

Analyzing everything at maximum depth. Ten games skimmed beat one game dissected for six hours. Depth of insight, not depth of search, is the goal.

How Our Tools Fit Your Analysis Routine

The workflow above maps directly onto ChessMind AI. The free analysis board is your Step 3 home: import a game, get the eval graph, step through the swings, and save the 1.5+ positions as study material — no account gymnastics, no paywall on the basics. If the game exists only on paper — a tournament scoresheet or a position from a book — scan it with your camera and it lands on the analysis board ready to work.

And when you want the monthly-review insight without waiting months, take the GM diagnostic: it examines your play across many games and positions at once and identifies your leak profile — the same pattern-finding a coach does with a stack of your scoresheets, compressed into one session. It is the fastest way I know to find out which chapter of your leak journal will be the thickest.

Frequently Asked Questions

Should I analyze my chess games with an engine?

Yes — but second, never first. Do your own pass, commit to written opinions, then use the engine to check them. The engine is a brilliant fact-checker and a terrible teacher: it tells you what was best but never why you didn't see it, and the "why" is the part that improves you.

How long should I spend analyzing one game?

Thirty to sixty minutes total for a serious game: 15–30 minutes of your own analysis, about 20 minutes with the engine, five minutes writing the journal entry. Beyond an hour you are usually polishing variations rather than learning. Better three games at 45 minutes than one game at 4 hours.

Is it worth analyzing blitz games?

Lightly. Skim them in batches for two things: openings where you repeatedly get bad positions, and recurring one-move oversights. Deep analysis of blitz wastes effort, because most blitz mistakes come from the clock, not from your understanding. Save the full four-step method for rapid and classical games.

Do engine depth and settings matter for my analysis?

At club level, almost not at all. Any modern engine at modest depth spots every tactic relevant to your improvement in seconds. Obsessing over depth 40 versus depth 25, or engine A versus engine B, is procrastination dressed as diligence — the bottleneck is your diagnosis of your own thinking, never the engine's strength.

How many games should I analyze per week?

Two or three, done properly, is a strong routine — enough to keep the leak journal filling while leaving time to actually fix the leaks it reveals. Analyze every classical or rapid game you play if your volume allows. If you play mostly online rapid, pick the two most instructive games of the week: usually your most painful loss and your messiest win.