I guess a brilliant move is a move hard to find and very good.
I guess the problem is that Stockfish value your move relatively its own "best" move and that necessarily not a brilliant move even if it's the best. If your opponent drops the Q the best move is to take it. That doesn't mean it's a brilliant move.
Even if there is only one move that wins and other moves loses that doesn't mean the move is hard to find.
Therefore it would be difficult decide if a move is brilliant based on comp-evaluations.
I agree with LM TonyRo.
The computer cannot be programmed to mark a move as a good move or brilliant move.
It is just a matter of opinion.
Maybe a programmer can formulate some properties, that a good and hard to find move has, that also can be formulated in a way that can be implemented. It somewhat of a challenge however. For example Q sacs that are correct and hard to find and piece sacs that are complicated and correct.
A possible way to implement this could be using search depth. For example, if a move for white has a score of -4 on depth 12, but a score of #14 on depth 13, this is a pretty hard move to find and probably deserves an exclamation mark.
Indeed interesting idea, Messy
If stockfish says ´+0.00 on depth 20 and +4.00 on depth 21 and position is complicated it is probably a brilliant move. :)
In lichess analysis, mistakes are defined as follows:
Inaccuracy: 50 centipawns loss
Mistake: 100 centipawns loss
Blunder: 300 centipawns loss
So why not use exclamation marks like:
! good move: 100 centipawn gain
!! excellent move: 300 centipawn gain
or something like it.
And what about '!?'? :-)
#17: Because in computer chess a move can't gain score.
The best you can hope for is to keep the centipawn score (since the score of current position is defined by best move from that position).
But as others have suggested there is a viable solution to defining what we mean by "!" in computer chess.
It's simple:
1) The move is the best one, or close to best, as far as Stockfish can tell
AND
2) A dumb search (e.g. limited depth and maybe some other defined dumbness) will label the move as strongly suboptimal. ("Strongly" just means that its score from shallow search is significantly below the best score from a shallow search).
In other words: It is the best, or close to best, move, and it takes a lot of effort to find it.
I think it would be a great annotation for multiple reasons. E.g. when reading a game, it helps finding the interesting moves - good moves which are not obvious.
Does anyone know how could someone start to implement the idea that trebe and others supported in order to provide the '!' annotation?