Move classifications label how much a move cost compared with the best move available. The usual grades are best, excellent, good, inaccuracy, mistake and blunder, ordered by the amount of advantage given away. They are derived from engine evaluations, so a label depends on the depth and settings used to produce it.
Move classifications are shorthand for how much a move cost. They are useful because they draw the eye to the handful of moments that decided the game, and dangerous because the labels sound more objective than they are.
The grades, in order
| Grade | Roughly means |
|---|---|
| Best | The move the engine would have played |
| Excellent | Nearly as good as the best move, no meaningful loss |
| Good | A sound move with a small cost that changes nothing |
| Inaccuracy | A minor concession, a few tenths of a pawn |
| Mistake | A real error, around a pawn or more |
| Blunder | A serious error, often several pawns or a lost position |
Some tools add a book grade for moves still inside known opening theory, and a missed win for a position that was winning and no longer is. Both are refinements of the same idea rather than separate scales.
How the label is actually derived
The usual method is centipawn loss: compare the evaluation after the best move with the evaluation after the move played, both from the mover's point of view, and measure the difference. A small difference is an excellent move, a large one is a blunder, and the thresholds in between define the other grades.
Two refinements matter. First, once a position is completely winning, further losses should not produce dramatic labels, since throwing away half of a plus nine position still leaves a plus four position. Second, in a position that is already lost, no move is really a blunder, because the damage was done earlier.
Why brilliant is not the same kind of label
The familiar brilliant marker is not simply a very good move. It implies something specific: a sacrifice that is sound, where the engine can prove the compensation works and that a plausible recapture fails. Establishing that properly needs a comparison of several lines at the position, not a single score, because the claim being made is that material was given up and the position is still winning.
This is why brilliant is the grade most often produced or skipped inconsistently. A false brilliant is worse than no brilliant at all, because it teaches a player to sacrifice for the wrong reason.
Using grades well
- Treat the labels as a reading order, not a scorecard. Start with the worst moves and ask what you missed.
- Ignore a single grade in isolation and look at the pattern. Repeated mistakes in the same phase point at something to study; one blunder in a long game may just be fatigue.
- Check the depth. A grade from a shallow search is provisional, and re-running at a higher depth can move it.
- Prefer the label that says what happened. A missed win tells you more than a generic blunder.
- Remember that accuracy percentages are summaries of these same scores, so they inherit the same caveats.