Why does the engine only look for mistakes and not for good moves?

@ProgrammerAngrim said in #10:
> One is that this is the only non-bad move
This illustrates nicely our point, actually. By your definition, if my opponent trades queens, the obvious recapture is a "good move". Or even worse, by this definition, even any forced move (i.e. the only legal move) is a "good move". Easy to implement, yes. Reasonable? I don't think so.

> The other is that this is a non-obvious good move, defined as a move that was not a top move after 4 ply, but is the best move after 12+ ply.
This is not so obviously flawed but not very useful either. There are moves that any strong player would call good or even brilliant (and far from obvious) that show as best even with calculation 4 plies deep. Or moves that look best 12 plies deep but still turn out to be bad when you look even deeper. As a rule of thumb, if your definition depends on a choice of constants that have to be completely arbitrary and cannot be based on any logical ground (4 and 12 here), it's going to be a problem.

Another important point is that even human players often won't agree if a move is "obvious" or not. Thus trying to find an exact definition or algorithm to determine it is IMHO futile.

@ProgrammerAngrim said in #10:
> It would actually be pretty easy for a computer to mark good moves, by either of two definitions. One is that this is the only non-bad move [...]

Not only it marks as good obvious moves (as noticed by @mkubecek ) but also fails to mark a good move when you find the absolutely best move among several good ones.

@ProgrammerAngrim said in #10:

To blow into the same horn as the previous answers:

> One is that this is the only non-bad move, singular moves are well defined and understood by computers.

This is actually simimilar to the John Nunn Convention (en.wikipedia.org/wiki/Chess_annotation_symbols#Nunn's_convention). While easy to implement, it seems dubious for practical analysis work.

Recaptures are an obvious miss. But simply ignoring them isn't correct either. And there are many obvious "only moves" that are trivial to spot and each beginner would play. Giving them an exclam would be pointless and devalue the whole annotation.

>The other is that this is a non-obvious good move, defined as a move that was not a top move after 4 ply, but is the best move after 12+ ply. Either one would be easy to code. Maybe you could call a move which is both singular and non-obvious a "!!" move

First, computer search is not as easy as a search tree at uniform depth. But the computer's calculation process is much different from human thinking. Some moves are trivial for humans, but the engine takes ages to find it, and obviously many examples the other way around.

If you guys think that this marketing-hyped-stuff at chess_com is so great, I suggest to have a deeper look there and examine how good it really is and how much value it brings...

One thing is for sure... complimenting the user for good moves makes them feel good, and they are happy, and happy users are more likely to spend money.

It reminds me of the fitness trackers... the one that shows the highest kcal burn are always perceived as best.

@AyaanshGaur12 said in #5:
> I get it, but if chess.com can do that, we can also do the same. Its not like we are living with neanderthals is it? Its the same 21st century.

Yes but does the definition they use make much sense. Opinions differ on this. Chess.com brilliancy is just move that sacs material which will be won back later - with interest. I.e you found combination. Which is fair is good amount cases but then what about finding the only non obvious winning move in endame.

So yes it can be done once developers would agree on what to do. And then some people would disagree and ask it to work differently. Hence I can understand why it has not been done.

Just to add another point to the "only winning move" concept:

Consider a position where one side has to play move A to make progress, but move B still wins (because after repeating the position he can still do the same move A). So move A might still warrant a "!", although not being the only move. Those things are hard to detect, and the positions arising after move B might not even be exactly repetitions, but just similar.

And then, what's a "!" move in a beginner's book might go completely uncommented on expert's level.

That's not saying, that it wouldn't be nice to have those annotations in a meaningful way (which humans can probably grasp intuitively quite well).

@petri999 said in #14:
> Yes but does the definition they use make much sense. Opinions differ on this. Chess.com brilliancy is just move that sacs material which will be won back later - with interest. I.e you found combination. Which is fair is good amount cases but then what about finding the only non obvious winning move in endame.
>
> So yes it can be done once developers would agree on what to do. And then some people would disagree and ask it to work differently. Hence I can understand why it has not been done.
Everyone knows abt brilliant moves (!!), but what abt great moves (!)? Also, its gonna be SUPER EASY to just make a Best Move, since its the best move on the engine. And I guess Excellent moves are the second line, and good moves are the 3rd line move(s).

@AyaanshGaur12 said in #16:
> Everyone knows abt brilliant moves (!!), but what abt great moves (!)? Also, its gonna be SUPER EASY to just make a Best Move, since its the best move on the engine. And I guess Excellent moves are the second line, and good moves are the 3rd line move(s).
so if the best move is to recapture a queen, but instead I sacrificed a rook with check which happens to be the only move that still allows me to take the queen next turn, is it an excellent move?

I swear half of these comments don't even try to think about the problems with their " obvious solutions". confirmation bias at it's finest

@GW-ISR said in #17
> so if the best move is to recapture a queen, but instead I sacrificed a rook with check which happens to be the only move that still allows me to take the queen next turn, is it an excellent move?
>
> I swear half of these comments don't even try to think about the problems with their " obvious solutions". confirmation bias at it's finest
Unless the rook sacrifice does something which will make you win the game after you recapture the queen, it would be marked either good or a dubious/mistake move. If it DOES indeed improve your position after recapturing it would be marked a "brilliant" move for sacrificing.

Lichess could show stockfish score for every moves it has calculated at each stage, that's a factual data and every one could interpret by himself. If say the best move as a far higher score than the second best one, then it can be interpreted as a good move.

@AyaanshGaur12 said in #18:
> @GW-ISR said in #17
>
> Unless the rook sacrifice does something which will make you win the game after you recapture the queen, it would be marked either good or a dubious/mistake move. If it DOES indeed improve your position after recapturing it would be marked a "brilliant" move for sacrificing.
according to your previous post it will always be marked as an excellent move, beacuse even if it just gives away a rook for free instead of recapturing, it is better to be down a rook then down a queen.