Trust & clarity

How Analysis Works

The short version of what ChessIQ measures, why labels can move, and how a reviewed move becomes a saved follow-up position.

What is ChessIQ actually measuring?

It compares your move with the best line Stockfish found, then tracks how much the position and expected result changed.

Are labels and accuracy the same thing?

No. Labels can use expected-score context; adjusted accuracy stays CPL-based and relevance-weighted.

What can ChessIQ save for follow-up?

Only suitable missed chances, conversion chances, and major mistakes become saved positions.

Move labels

ChessIQ looks at the position before your move, the position after your move, and the best continuation Stockfish found. Most labels start from centipawn loss, then severe labels can be softened by expected-score loss when the practical result did not change as much as the raw evaluation swing suggests.

Forced moves, book moves, mates, low-confidence positions, and ambiguous top lines have extra guards. Some labels can stay provisional until the depth, confidence, and played-move comparison are strong enough.

Accuracy and ChessIQ

Adjusted accuracy is still CPL-based. It uses centipawn loss, score provenance, and a relevance-weighted averaging pass so routine or already-decided moves do not dominate the summary.

ChessIQ profile scores summarize local reviewed history: accuracy, results, color balance, openings, time controls, opponent context, and stability/volatility signals. Confidence rises with sample size and coverage.

Saved follow-up positions

A move does not become a saved position just because it has a scary label. ChessIQ checks that it was your move, that the score evidence is real, that the best move is useful, and that the position passes validation.

Saved-position priority can include blunders, major mistakes, missed opportunities, and conversion chances. It also uses local attempt history, due timing, and recurrence so the same weak spot can come back until it stabilizes.

Where analysis runs and where data lives

Core analysis runs in your browser using Stockfish WebAssembly. Core review history, saved-position state, and many preferences are stored locally (including IndexedDB/local storage paths used by the app).

External calls are still used for public-game imports (Chess.com/Lichess) and production analytics, so local-first should be read as a trust boundary - not an absolute "no network ever" claim.

Limits, caveats, and edge cases

  • Deeper analysis can legitimately change evaluations, best lines, and labels.
  • Device and browser performance affect review speed and how quickly results stabilize.
  • Chess960 is recognized but has limited downstream support in saved positions/statistics.
  • Unsupported variants should be rejected rather than partially interpreted.
  • ChessIQ explains review evidence; it does not claim to measure chess talent or player potential.