Analysis Is Not Finished Until Something Changes
A beautifully annotated game can still produce no improvement if its findings never become memory, training or behavior. The final stage is compression: turn dozens of variations into a small number of reusable outputs.
Separate Game-Specific Corrections From General Lessons
“On move 27 I should have played ...c5” is game-specific. “When my opponent fixes the queenside, check the central break before defending passively” is reusable. Keep both, but do not confuse them. The second statement is what transfers to future positions.
Match the Output to the Root Cause
Different diagnoses imply different training:
- missed tactical motif → targeted tactical set;
- candidate-generation failure → candidate drills;
- visualization error → branch-state work;
- endpoint mis-evaluation → comparative evaluation positions;
- opening gap → repertoire note + model game;
- endgame knowledge gap → theoretical position recall;
- clock failure → time-allocation rule or simulated time drill;
- execution error → interface/procedural practice.
Training should repair the cause, not merely replay the punishment.
Build a Recurrence Log
Across games, count themes and causes. One error can be accidental; a repeated pattern deserves priority. A simple log with date, position/theme, root cause and training action is enough to expose recurring weaknesses without turning improvement into bureaucracy.
Save Model Positions
When a game contains a clean example of a concept—good exchange, key pawn break, defensive resource, endgame transition—save the position with a one-sentence lesson. Over time, your own games become a personalized model-position library.
Close the Loop
The strongest feedback loop is:
play → preserve → analyze → diagnose → train → replay similar decisions → reassess.
The goal is not to accumulate engine-approved games. It is to make the next independent decision better.
Prioritize Lessons by Frequency, Cost and Trainability
A useful improvement queue considers three dimensions: how often the error appears, how costly it tends to be, and whether a concrete training method exists. A rare catastrophic mouse slip may need an execution habit; a frequent moderate candidate-generation failure may deserve much more training time.
Test Whether the Lesson Is Too Broad
“Calculate better,” “improve strategy,” and “manage time” are not yet training instructions. Compress the lesson until it names an observable behavior: “before every recapture, scan checks and intermediate moves,” or “when leaving preparation, spend one deliberate reset to evaluate the current position.” Specific lessons are easier to rehearse and later audit.
Reference Framework
The final product of analysis is not a verdict on the past. It is a reusable improvement asset.
A complete report moves:
game → critical moment → evidence → root cause → correction → general lesson → repertoire/training action
Game-Specific Correction
A game-specific correction states what should have happened in that particular position and why.
Reusable Chess Lesson
A reusable chess lesson generalizes the cause: before a pawn break, check the resulting open file, not merely play 23...Re8.
Repertoire Update
A repertoire update changes your maintained opening notes when analysis identifies a move-order gap, theoretical problem or better response.
Training Task
A training task turns the root cause into deliberate practice: candidate-generation set, visualization drill, calculation line, endgame position, strategic comparison or clock exercise.
Model Position
A model position stores a clean FEN/example representing the transferable strategic/tactical idea.
Endgame Position
An endgame position stores a theoretical/practical reference with exact result and human-method verification when available.
Practical / Clock Lesson
A practical/clock lesson stores the decision-process correction from practical-play guidance: criticality, overspend, autopilot, risk or recovery.
Repeated Tactical Error
Repeated tactical errors across games indicate a recurring recognition/calculation issue rather than isolated bad luck.
Repeated Candidate Failure
Repeated candidate failures show a search-generation bias—missing quiet moves, defensive resources, sacrifices, counter-checks or transformations.
Repeated Structural Misread
Repeated structural misreads show that the same pawn structure/weakness/break is being evaluated or planned incorrectly.
Opening Gap
Repeated opening gaps reveal repertoire branches, transpositions or typical-position knowledge that need maintenance.
Endgame Gap
Repeated endgame gaps identify exact theory/technique families for focused drills.
Time-Management Pattern
A time-management pattern is recurring over/underspending, no reserve, repeated calculation or poor critical-moment recognition.
Practical-Risk Pattern
A practical-risk pattern is recurring misuse of complexity, simplification, robustness, only-move dependence or result context.
Metadata / Context
Analysis report metadata/context should identify game/source/time control and enough provenance to reproduce the review.
Critical Moments
Critical moments section lists the decisions worth deeper attention rather than annotating every move equally.
Human Thought
Human thought section preserves candidates, expected replies, believed evaluation, plans and uncertainty.
Tool Findings
Tool findings separate database, engine and tablebase evidence rather than blending all computer output together.
Root Causes
Root causes state trainable reasons, not only consequence labels.
Repertoire Changes
Repertoire changes record exact position/move/source/date so they can be maintained later.
Training Actions
Training actions are specific enough to execute: task type, positions/material, target skill and review loop.
Short Summary
Short summary should answer: biggest chess lesson, biggest process lesson and one concrete next action.
Analyze Without Engine First
Mastery requires analyzing without engine first often enough to preserve independent reasoning ability.
Compare with MultiPV
Mastery requires comparing recorded human candidates with MultiPV without mistaking MultiPV for absolute truth.
Audit Opening Database
Mastery requires auditing opening database results with filters/sample/transpositions rather than copying explorer percentages.
Verify Endgame with Tablebase
Mastery requires exact endgame verification with tablebase when applicable and separate diagnosis of the human error.
Annotate a Game
Mastery requires annotating a full game with clear played line, selected variations, explanations and provenance.
Extract Three Reusable Lessons
Extract at least three reusable lessons only when the game actually supports them; do not manufacture general rules from one accident.
Analysis Mastery Set
The analysis mastery set should mix human-first analysis, root cause, CP/WDL/PV interpretation, MultiPV, database filters, transpositions, tablebase and annotation tasks.
Readiness for Training & Improvement
Readiness for training guidance means you can turn analyzed games into validated training targets rather than collecting engine verdicts.
Next: training guidance — Training & Improvement