Analysis Needs Both Narrative and Reproducibility
A good annotation record lets a future reader reconstruct the game, locate the exact position, understand the claim and see why a variation was included. Raw PGN, annotated PGN and critical-position FENs serve different purposes and should not be conflated.
Preserve the Raw Game
Keep an immutable original when possible. Work on an annotated copy. That protects source tags, clock comments, termination data and provenance from accidental editing and makes later corrections auditable.
Use FEN as a Position Anchor
For critical moments, a FEN provides a precise root independent of the surrounding movetext. It is especially useful for engine reruns, tablebase probes, diagrams and training cards. It does not replace the full game record because it cannot preserve every historical detail of how the position was reached.
Variations Must Earn Their Space
Include a line when it demonstrates a refutation, best defense, alternative plan or exact transition. Avoid dumping every engine branch. Stop when the reader can understand the mechanism. A short line plus a clear explanation is often more valuable than twenty uncommented moves.
Keep Evaluation Types Distinct
Human symbols such as += or ∞, prose evaluation, numeric engine scores and tablebase WDL are different forms of annotation. Use them intentionally and record perspective/configuration when a number matters.
Provenance Makes Analysis Durable
For tool-dependent claims, useful provenance can include engine/version, analysis date, search limit, MultiPV, relevant tablebase configuration, database/corpus, filters and sample size. Not every casual note needs all fields, but stronger claims require stronger records.
Keep Machine-Readable Tokens Stable
When analysis is localized or reformatted, PGN tags, FEN strings, SAN moves, engine options and other machine-readable tokens should remain exact. Translate the surrounding explanation, not the data that software must parse. This distinction becomes especially important when the same analysis is reused across web, books and training tools.
Comments Should Explain the Decision Boundary
Strong annotations answer why one move is preferable to another: what threat is met, which square becomes available, what ending is reached, or which defender is overloaded. Weak annotations merely restate the move or attach a score. A useful comment often survives even if a future engine version changes the exact numerical evaluation.
Version Derived Records
If an annotated PGN, repertoire file or analysis notebook is edited repeatedly, preserve enough version history to distinguish raw evidence from later interpretation. The goal is not bureaucratic version control; it is avoiding a situation where today's corrected annotation is mistaken for what you knew immediately after the game.
Reference Framework
Analysis must remain auditable. The raw PGN is evidence; annotations, critical-position FENs, repertoire notes and training notes are derived records.
A durable analysis record should let a future editor reproduce both the chess position and the tool claim.
Raw PGN
Keep the raw PGN immutable or separately preserved as the source record.
Annotated PGN
Annotated PGN is a derived working/documentation artifact containing comments, variations, NAGs and possibly tool metadata.
Critical-Position FEN
Store critical-position FENs when a position must be referenced independently of the game move sequence.
Repertoire Note
A repertoire note records the position, chosen response, rationale, alternatives, sources and date/version of the repertoire decision.
Training Note
A training note converts an error into a drill/task with trigger, target skill and success criterion.
Played Line
The played line must remain visually distinguishable from analytical variations so readers never confuse analysis with moves actually played.
Main Analytical Variation
The main analytical variation demonstrates the principal correction or best-defense line relevant to the lesson.
Side Variation
Side variations answer important alternatives without overwhelming the main explanation.
Comments
Comments explain reasons, plans, errors, uncertainty, source context and lesson extraction.
NAGs
NAGs are compact annotation symbols defined by the PGN/notation conventions. They should support—not replace—explanation.
Evaluation Symbols
Evaluation symbols such as += or ∞ are human/editorial summaries whose exact conventions should remain consistent within the publication.
Numeric Engine Evaluation
Numeric engine evaluations require perspective, engine/version/search context when precision matters. A bare +1.3 is weak provenance.
Avoid Variation Dumping
Avoid variation dumping: ten engine lines without explaining the critical idea usually hide rather than reveal the lesson.
Explain Why
Explain why a move works/fails: tactic, defender, line, structure, king safety, endgame transition or plan—not just engine prefers ....
Player/Event/Date/Round
Record player/event/date/round for provenance and database matching.
Source Database
Record the source database/game source and, where relevant, snapshot/version because records can be corrected later.
Annotation Author
Record annotation author/source so human interpretation is attributable.
Engine Version
Record engine version because evaluation/search behavior changes across releases.
Search Settings
Record search settings when reproducibility matters: MultiPV, limit, threads/hash if relevant, tablebase configuration and other materially relevant options.
Analysis Date
Record analysis date because engines, databases and theory evolve.
Database Snapshot
Record database snapshot/date and filters for statistical/theory claims.
Version the Analysis
Version the analysis when corrections, deeper searches or theory updates materially alter conclusions. Do not silently overwrite historical analytical states.
Annotation Practice
For annotation practice, transform raw engine or database output into a concise human explanation with provenance and one reusable lesson.