Brilliant Move in Chess: What Counts and What Does Not

Why there is no shared definition

Chess has no rulebook for “brilliant”. Players use the word for a beautiful sacrifice. Review tools use it for a label, and each tool decides when to give it. So a brilliant on one site may be a plain best move on another. This page explains only our rule, which you can verify.

Our rule for a brilliant move

A move is brilliant only when all of these are true:

• The deeper engine pass checked it, both before and after the move.
• It is the engine’s best move, or it loses less than 2 Win% points.
• It is a sacrifice: it leaves at least 2 points of material to be won, and the opponent’s best reply does not simply give it back.
• After the move you have between 30% and 97% winning chances.
• Before the move you had at least 15%, so it is not a desperate swindle.

Material counts as pawn 1, knight 3, bishop 3, rook 5, queen 9. Sacrifices of material that was already lost do not count.

Great versus brilliant

A great move is the engine’s best move, and the second-best option is at least 15 Win% points worse. It is the only move that holds the position or keeps the win. It does not have to be a sacrifice. Recaptures and moves that answer a check never count, because they are obvious.

A brilliant move needs a sacrifice and a Win% that still has room to grow. That is why a spectacular sacrifice in a position that is already winning by force is labelled great, not brilliant.

How rare is it?

Across the 12 games we ran through the engine, 584 plies in total, 2 moves were brilliant. Both came from historic games. We found 31 great moves in the same set. Treat it as a small sample, not a statistic about chess in general. The point is that the label is rare on purpose: if every sacrifice earned it, it would tell you nothing.

A real example: Lasker against Bauer, 1889

In the engine’s review of this game, 17.Bxg7 is brilliant. It goes from 81.4% to 82.4%, so the sacrifice costs nothing in winning chances.

Open the line on the board and you can check it yourself. That is the standard we want for every label.

Find yours

Most games will not have one, and that is normal. Run a game through the free analysis and look at the key moments, or read how the checker measures errors on the blunder checker page. For a set of games, the report shows patterns.