Preserve Evidence Before Hindsight Rewrites It
Post-game memory is fragile. Once you see a strong engine move, it becomes surprisingly easy to remember that move as something you had “almost considered.” Once you learn the evaluation, your earlier uncertainty can disappear from memory. For improvement, that lost uncertainty is valuable evidence.
Record context before heavy tool use whenever practical: where theory ended, what surprised you, candidate moves, expected replies, the evaluation you believed, the plan you intended, and whether you felt confident or confused.
Reconstruct the Decision, Not a Fictional Monologue
You do not need a perfect transcript of your thoughts. In fact, pretending to remember every variation can create false precision. A good reconstruction distinguishes:
- remembered fact — “I considered 18...c5.”
- approximate memory — “I think I expected a queen trade.”
- unknown — “I do not remember evaluating the rook ending.”
- inference — “Given my next move, I probably underestimated the pin.”
Marking uncertainty makes the record more credible and more useful.
Clock Context Changes the Diagnosis
The same move can have a different root cause under different clock states. Missing a three-move tactic with twenty minutes remaining may indicate candidate or calculation failure; missing it with three seconds may instead point to earlier time allocation. Exact clock values should come from reliable records when available. If you only remember “I was very low on time,” store that qualitative fact rather than inventing a timestamp.
Capture the Model You Were Using
The most diagnostic notes often describe assumptions rather than variations:
- “I thought the queenside was closed.”
- “I believed the exchange favored me.”
- “I assumed the knight was trapped.”
- “I thought this was still my preparation.”
These beliefs tell you what internal model produced the move. Later analysis can test the model itself instead of merely replacing the move.
Raw Record and Working Copy
Keep the original game immutable. Analyze a copy. This protects source metadata, clock comments, annotations and provenance from accidental overwriting and lets you distinguish what existed at game time from what was added later.
Capture Emotional and Practical Context Carefully
Practical state can matter: surprise, frustration, confidence, fear of a line or a desire to simplify may explain why a candidate was ignored. Record these as decision context without turning every error into psychology. The board still needs chess analysis; context helps explain why your process departed from it.
Separate What You Saw From What You Calculated
“I noticed the back-rank weakness” is not the same as “I calculated the forcing line.” Distinguish recognition, candidate generation and verified calculation. This prevents analysis from crediting pattern recognition with a variation you never actually worked through.
Reference Framework
The raw game and the player's remembered thought process are evidence. They should be preserved before later analysis replaces uncertain memories with hindsight.
Record what you actually thought—including uncertainty—not what the engine later makes you wish you had thought.
Preserve Raw PGN
Preserve the raw PGN unchanged whenever possible. The original record is evidence; annotations belong in a working copy or versioned derivative.
Annotated Working Copy
Use an annotated working copy for comments, variations, NAGs, engine metadata and corrected/expanded analysis while retaining the raw source separately.
Players / Event / Date / Round
Preserve players, event, date and round because identity/context affect provenance, historical research and database matching.
Result / Termination
Preserve result and termination separately. 0-1 does not tell whether the game ended by mate, resignation, time or another mechanism.
Time Control
Preserve the stated time control because it materially changes the practical interpretation of decisions.
Clock Data
Preserve clock data when reliably available. Do not invent exact think times from memory, edited video or delayed commentary.
Original Source / Game ID
Record original source/game ID so future corrections, duplicates or richer metadata can be traced.
Opening Metadata
Opening metadata can be recorded, but position/FEN remains more fundamental than an opening label when transpositions occur.
Where Did Theory End?
Ask where your reliable theory memory ended. This is not necessarily where published theory ends.
Where Were You Surprised?
Record positions where the opponent surprised you because surprise changes practical decision-making even when the move is theoretically known.
Candidates Considered
Record candidates actually considered before engine reveal. Missing the best move can then be diagnosed as candidate generation rather than only bad calculation.
Expected Replies
Record expected opponent replies. A calculation failure may arise because the opponent's strongest defense never entered the tree.
Evaluation Believed
Record the evaluation you believed—better/equal/worse and why—without rewriting it after seeing the engine.
Plans Considered
Record plans considered so strategic-plan failures can be distinguished from execution errors.
Clock Spent
Record clock spent only when reliable data exists; approximate self-report should be labeled approximate.
Move Played Too Quickly
Mark moves played unusually quickly relative to criticality. The root cause may be autopilot or time allocation rather than chess ignorance.
Tactical Ideas Seen / Missed
Separate tactical ideas seen from tactical ideas missed. I saw the motif but miscalculated it differs from the motif never entered my candidate set.
Preserve Uncertainty
Preserve uncertainty explicitly: I considered A/B; I was unsure about .... Uncertainty is analytical evidence, not something to clean away.
Memory Decays
Memory decays and hindsight contaminates reconstruction. The sooner the thought record is made, the more trustworthy it usually is.
Reconstruction Practice
For reconstruction practice, rebuild several critical decisions from PGN, clock and context before revealing any engine labels.