Every move you play gets a label. Here is exactly what each one means and where the line is drawn.
Every move you play during analysis gets compared against Stockfish's own best move at that position, and the gap between them — called centipawn loss — decides the label. A centipawn is 1/100th of a pawn, so a loss of 100 centipawns is roughly a full pawn's worth of accuracy given up. Here's the exact breakdown:
Book — A well-known opening move, matched against real games in the Lichess opening database — not centipawn-based at all.
Brilliant — A sacrifice that is also the engine's best move, gains 200+ centipawns, and isn't played from an already-crushing position. Rare by design.
Great — Either a strong sacrifice that holds the position, or a non-sacrifice best move that rescues a losing or equal position into a clearly winning one.
Best — Zero centipawn loss — this was the engine's own top choice.
Excellent — Under 10 centipawns lost. Effectively as good as the best move in practice.
Good — 10–24 centipawns lost. A solid, reasonable move.
Inaccuracy — 25–99 centipawns lost. Not dangerous on its own, but worth noticing.
Mistake — 100–299 centipawns lost. A meaningful chunk of advantage given up.
Blunder — 300+ centipawns lost. A serious drop — often the moment a game turns.
Miss — A special case: you were clearly winning (150+ centipawns ahead), lost over 100 centipawns of it, and ended up no longer clearly winning. This flags exactly the "opponent handed you a win and you let it slip" moment.
"Best" strictly means you matched the engine's own top choice for that exact position — by definition, zero loss. But it's worth remembering the engine's best move is only ever the best move according to its own search, at whatever depth it ran. It isn't necessarily the most human, most practical, or safest choice against a real opponent — just the most accurate one on paper.
Brilliant specifically requires a sacrifice — giving up material — that turns out to also be the objectively correct move and that gains a large amount of evaluation. Most good moves in a normal game are simply "Best" or "Excellent." Brilliant is reserved for the genuinely striking, memorable moments, the same way it's rare in a real game to have a chance to sacrifice a piece and have it actually work.
Inaccuracies are usually not worth agonizing over — they're common even in strong games. Mistakes and Blunders are the ones worth slowing down on: pull up that exact position on Explore and see what the engine actually preferred, and why. A Miss is often the most instructive of all, since it means the position was already yours to win — the question isn't "what should I have played instead," it's "what did I overlook."
Want to see how this plays into the evaluation number itself? Read How to Read a Chess Engine Evaluation, or head straight to Analyze to try it on one of your own games.