Most players already know they should “analyze their games.” The advice is so common it has become useless. You open a tab, click Game Review, scroll through the bouncing evaluation bar, nod at the mistakes Stockfish flags, and close the tab. A week later your rating has not moved.
The problem is not that you are lazy. The problem is that engine-driven game review is a verification tool, not a diagnostic method. It tells you where you went wrong; it does not tell you why you keep going wrong in the same way. After coaching club players from 1100 to 2100, I have found that almost every plateau is caused by one of five recurring decision errors — not by gaps in opening theory or missed tactics in the moment.
This article gives you a five-pass diagnostic framework I use with students. It is designed to surface patterns across many games rather than chasing single-move blunders. By the end, you will know exactly what to study next — and why.
Why Standard Engine Review Fails
An engine evaluates positions; it does not evaluate decisions. When Stockfish flags move 23 as a “?” and suggests an exchange sacrifice you never would have found, the engine is correct about the position and useless about your improvement. You did not lose that game because you missed a +1.4 exchange sacrifice. You lost it because you committed to an attacking plan three moves earlier without checking whether the defender had a stable structure.
That earlier moment — the one with no “?” annotation — is where your rating lives. The engine cannot see it because the engine has no model of you. For more on this trade-off, see our breakdown of Stockfish analysis vs a human coach.
The Five-Pass Diagnostic Method
Run every serious game (classical, rapid 15+10, or longer) through these five passes. Skip blitz; the time pressure introduces noise that drowns out the signal. Each pass takes 6–10 minutes once you are used to the system, so a full analysis runs about 30–40 minutes — far less than the hour most players waste on engine-only review.
Pass 1 — The Memory Pass (no engine)
Replay the game from move 1 without an engine open. At every move, write a one-line note: what you were thinking, what you feared, and which candidate moves you considered. If you cannot remember, write “no plan.” That answer is the most useful diagnostic data you will produce all week.
The Memory Pass forces you to separate positional understanding from engine-aided hindsight. A pattern of “no plan” notes between moves 15 and 25 is the classic middlegame-drift fingerprint. You are not losing because you blunder; you are losing because you stop having opinions.
Pass 2 — The Critical Moments Pass
Now turn on a low-depth engine (depth 18–22 is plenty; the deeper analysis is noise for human improvement). Mark every move where the evaluation swings by 0.8 or more. These are your critical moments. Ignore everything else.
For each critical moment, label it with one of five tags:
- Calculation error — you saw the right idea but miscounted a line
- Evaluation error — you reached the end of a line and judged the resulting position wrongly
- Candidate error — the right move never entered your list of options
- Time error — you knew the answer but were rushed or burning clock
- Plan error — the local move was fine but served a broken long-term plan
These five tags are deliberately mutually exclusive. Forcing yourself to pick one is the entire point. If you find yourself wanting to tag a move “candidate + calculation,” your tagging muscle is not yet developed — pick the earlier root cause.
Pass 3 — The Pattern Pass (across games)
Single-game analysis tells you almost nothing. The diagnostic power lives in cross-game patterns. After tagging 10 games, count your tags. The distribution will look something like this:
- Candidate errors: 14
- Plan errors: 11
- Calculation errors: 6
- Evaluation errors: 4
- Time errors: 3
This player does not need a tactics course; they need a candidate-move discipline (the look-wider-before-deeper habit) and a planning framework. A different player with 14 calculation errors and 2 candidate errors needs the opposite — visualization drills and a structured way to calculate variations cleanly.
The point is that your tag distribution prescribes your study, not the other way around. Most players have it backwards: they pick training material based on what is fashionable or what their favorite YouTuber covered last week, and the training touches none of their actual leak points.
Pass 4 — The Archetype Pass
Now zoom out further. Across your last 10 games, which kinds of positions did you mishandle?
Sort your losses into three buckets:
- Sharp, open positions with king safety and tactics dominant
- Closed, maneuvering positions with pawn-structure decisions dominant
- Endgames where technique and conversion dominate
A player who loses 7 of 10 in sharp positions but wins maneuvering games is not a “calculator who needs more tactics” — they are a positional player whose opening repertoire forces them into sharp lines they cannot defend. The fix is usually a repertoire change, not 1,000 more puzzles. This is why we think about improvement in terms of chess archetypes: your archetype determines which training transfers and which is wasted.
Pass 5 — The Decision-Tree Pass
For one or two of your worst games, build a decision tree at the most pivotal moment. Write out the three candidate moves you considered, why you rejected two of them, and what you believed the third would achieve. Then compare to the engine’s top three options and notice where your decision tree diverged from the correct one.
Most players discover, repeatedly, that the right move was in their original candidate set — they rejected it because of a single concrete line they miscalculated, or because of a fuzzy “this feels bad” intuition that turned out to be wrong. This is gold. Intuitions you can name are intuitions you can retrain.
How to Use the Diagnostic Output
After running this method on 10–15 games, you will have three pieces of data: your tag distribution, your archetype loss profile, and a small library of decision trees. Together they tell you what to do next:
- If candidate errors dominate, you need a candidate-move protocol (e.g., the “list three before calculating any” rule from Kotov, adapted for online time controls).
- If plan errors dominate, you need pawn-structure study tied to the actual structures you reach from your openings — not generic middlegame books.
- If calculation errors dominate, you need short, daily visualization work; long puzzle sets are mostly noise.
- If time errors dominate, you have a clock-management problem, not a chess problem — see our piece on chess time management at every rating.
- If evaluation errors dominate, you need to study annotated master games slowly, predicting moves and explaining your evaluation before turning the page.
Notice that none of these prescriptions is “do more of everything.” Improvement is not a volume problem. It is a targeting problem.
Common Mistakes When Self-Analyzing
Three failure modes show up reliably when players try to run this method on their own.
Tagging is too generous. Almost every player initially under-counts plan errors and over-counts calculation errors, because plan errors are uncomfortable to admit. If your distribution shows zero plan errors across 10 games, your tagging is wrong, not your play.
Engine depth becomes a crutch. Running Stockfish at depth 40 to “verify” your analysis defeats the purpose. The diagnostic value is in your process, not in the engine’s evaluation. Cap the engine at depth 22 and move on.
The Memory Pass gets skipped. It is the most boring pass and the highest-leverage one. The whole framework collapses without it, because you lose access to what you were actually thinking during the game.
Where the $14.99 Plan Fits
If you would rather not run this manually, the MyChessPlan personalized improvement plan automates the tag distribution and archetype profile for you — you upload 10 games and receive a written diagnosis and a 4-week training plan calibrated to your specific leak points. It is the same five-pass logic above, but the bookkeeping is done for you and the training prescriptions are pulled from a structured library rather than rebuilt from scratch each time.
You can also start with the free archetype report, which gives you the Pass-4 information by itself. That alone is often enough to reorganize a stagnant training routine.
The Bottom Line
Stop running engine review as if it were analysis. Engine review verifies; this five-pass method diagnoses. Diagnosis is what unlocks targeted study, and targeted study is what moves rating. Most players are not under-trained — they are mistrained. The fix starts with knowing exactly which of the five errors is bleeding the most points out of your game.
Run the method on your next 10 serious games. Write the tags down. Then come back and look at the distribution. The training plan you need is the one the distribution writes for you.
Frequently Asked Questions
How many games should I analyze with this method?
Ten games is the minimum for reliable tag distribution; 15–20 is ideal. Fewer than 10 produces too much noise for the patterns to emerge. Use only classical or rapid (15+10 or longer) — blitz noise corrupts the data.
Can I use Chess.com Game Review or Lichess analysis for this?
Yes, but only for Pass 2 (Critical Moments) onward. Pass 1 (Memory) must be done with the engine off. Both platforms work; for a comparison of their analysis features see our piece on Lichess vs Chess.com analysis.
How long does the full method take per game?
About 30–40 minutes once you are practiced. The Memory Pass is 6–10 minutes, Critical Moments 8–12, Pattern is essentially free (you tally tags across many games), Archetype takes 5 minutes, Decision-Tree takes 10–15 for the games you choose to deep-dive.
Do I need a coach to do this?
No. The method was specifically designed for self-analysis. A coach accelerates Pass 4 and Pass 5 because they have seen the patterns before, but most players can run Passes 1–3 alone after a couple of attempts. If you would prefer an automated diagnosis, the MyChessPlan personalized plan does the tagging and archetype work for you.
Leave a Reply