Its so depressing reading what that stupid thing ALWAYS has to say.
It never gives you a !, a !! or even a !?
Why cant they programme it to say something like "good move" or even " yes that is the move I would have recommended ?
L xxx
Why don't teachers mark the correct answers instead of the wrong answers? Why don't teachers send a list of students who are present instead of the ones who are absent? The computer assumes that you make more good moves than bad, so it marks the way that would prove to be the least amount of work.
@1
Write it at forum/lichess feedback it is a good feature for analyzed game.
There are easy moves that do not need an annotation but there are moves that the engine find the best move at higher depths only and it could not find it at lower depths that move could have a good or very good or interesting annotation symbol.
The machine cannot understand which move is good, excellent or interesting.But it can understand which move is a mistake ( most of the time anyway :), and I have to agree it is pretty ruthless
The machine can understand which is good and which is excellent or best move, it depends on the programmer, but there are methods of showing it.
Simple illustration, run the engine in multpv 3 for example.
FEN: r1bq1rk1/2p1bppp/p1n2n2/1p1pp3/4P3/1BP2N2/PP1P1PPP/RNBQR1K1 w - d6 0 9
Stockfish_16011823_x64_modern:
29/50 02:48 262,046,403 1,554,691 -0.13 d2-d3 d5xe4 d3xe4 Qd8xd1 Bb3xd1 h7-h6 Nb1-d2 Bc8-e6 a2-a4 b5-b4 Bd1-e2 Rf8-d8 Be2-c4 Be6-g4 Nd2-b3 Rd8-d6 Nb3-d2 Bg4-h5 Nd2-b3 Be7-f8 h2-h3 Rd6-d8 Bc1-d2 Bh5-g6 c3xb4 Bg6xe4 Bd2-c3 Be4xf3 g2xf3 Bf8xb4 Bc3xb4 Nc6xb4 Re1xe5 Nb4-c2 Ra1-c1 Nc2-b4
29/50 02:48 262,046,403 1,554,691 0.00 d2-d4 e5xd4 e4-e5 Nf6-e4 c3xd4 Bc8-f5 Nb1-c3 Ne4xc3 b2xc3 Nc6-a5 Bb3-c2 Bf5xc2 Qd1xc2 Na5-c4 a2-a4 c7-c6 Nf3-d2 Qd8-c7 Nd2xc4 b5xc4 Bc1-e3 Ra8-b8 Re1-b1 c6-c5 f2-f4 Rf8-e8 d4xc5 Rb8xb1+ Ra1xb1 Be7xc5 Be3xc5 Qc7xc5+ Qc2-f2 Qc5-a3 Qf2-c2 Qa3-c5+
29/50 02:48 262,046,403 1,554,691 +0.22 e4xd5 Nf6xd5 Nf3xe5 Nc6xe5 Re1xe5 c7-c6 d2-d4 Be7-d6 Re5-e1 Qd8-h4 g2-g3 Qh4-h3 Bb3xd5 c6xd5 Qd1-f3 Bc8-g4 Qf3xd5 Ra8-d8 Qd5-g2 Qh3-h5 h2-h4 Bg4-f3 Qg2-h2 Bf3-d5 Bc1-g5 f7-f6 Bg5-e3 Qh5-f5 h4-h5 Rf8-e8 h5-h6 g7-g6 a2-a4 Qf5-d7 Nb1-d2 b5xa4
At depth 29
best move: e4xd5 {+0.22}
good move: d2d4 {0.0}
The score gap is 0.22 - 0.0 = +0.22
An excellent move can be defined by a given gap margin, the score of the 2nd best move and at what depth the best move was found, does e4xd5 found at lower depths or it was found at depths 25 or more? If the best move e4xd5 is found at depths < 10, and it is still best at depth 29 right now, then this can be an easy move and could not be considered as excellent move. However if such best move was found at higher depths and it was not considered at lower depths, then this move can be considered an excellent move.
For interesting move there are conditions that should be satisfied, like its absolute score, its move score progression from lower depths to higher depths, phase of the game, position complexity and others.
The gap margin is irrelevant.I mean if you have 2 moves and you get mated in one and not in the other the gap margin would be very large but noone will be going as far as to say that because you found the forced move it was an excellent one.
The other idea is more decent but it has many problems, many of them which are technical, but lets assume you solve these(which are not trivial btw)
Now, for example, the machine considers the best move at 10 plies to be 1.e4 and at 25 plies it thinks that 1.d4 is better... You know where this is going I m too bored to explain it more
you could argue that not being mated is pretty excellent (it's just trivial) :p
now well, i think we should just embrace the ruthlessness. if you can't handle the truth (OOOHHH) then don't ask for the computer's rude opinions :o
@6
I am just showing you that a machine knows which move is good, easy, difficult, bad, excellent, which side has good piece mobility, king safety, threats, you name it, it is just a matter of manipulation, parameter definition etc.
It just takes time to analyze such game because you have to run the thing in multipv, do some calculations based on the output of the engine in order to get what you wanted.
You don't need to limit yourself to first moves, there are a lot of interesting positions out there that a multipv analysis would show to humans the reasons behind some moves.
It's not how good a move is (to the computer in every situation there is the best move, and bad moves to various degrees), it's how difficult it is to find. That's what we mean when we congratulate someone about a move. And I don't think the engine can know that. They can tell how difficult to find it is for them (based on how long it took them or how deep they had to calculate to find it), but since we don't think in the same way, this is not reliable for us. To be able to congratulate us, the engine should modelize our thinking. Also "difficult to find" only makes sense relatively, in the improvement path of a player. So they should not only understand how we think, but represent our evolution, and notice when we succeed in seeing something that we previously used to miss. A coach will provide you with exactly that, but an engine can't.
On the other hand with the computer you can auto-congratulate yourself every time you're in a situation where: there are a lot of tempting candidate moves, you think deeply, the move you finally select is the best move or one of the best moves according to the engine. If that happens regularly you are probably improving. I guess a very crude way to use this fact would be to congratulate you every time you take a long time to move and your move is one of the best ones...
We congratulate a human because he finds the best move, which according to engine is also the best move. That best move is not easy to find, not a simple recapture for example, not a simple mate in one, that best move is something that needs deeper calculation from human (well it depends on who this human is, as each of us has level of difficulties), which mostly like also needs deeper analysis depths from the engine.
There are root positions that changes the best move of an engine as it iterates thru the depths, at depth 1 it suggests Nxf6, at depth 6 it suggests Bc6, the more it changes its best move as it goes deeper, the more difficult the position to analyze is, at least for engine. At higher depths the engine analysis is more reliable, and if that best move is the same as what the human had moved then that move deserves ! or !! marks.