pichy wrote:PS: Notice that Komodo think like that only when you give the Black side to play, but if you give it the White side it will find the correct underpromotion to a Rook. I would like to see the Analysis of Don Dailey experimental program (to be released soon) which he already fixed.
That is obviously not true. In Uri's post Komodo is clearly thinking for white, and it does not find the under-promotion, but on the contrary wants to promote to Queen (and thinks it is won: +2.28 or #3).
Note that in my post I did not play a game against komodo from the initial position of this thread and it is possible that the behaviour of komodo3 in a game is different(meaning it see that promotion to a queen leads to a stalemate earlier and remember it in the hash tables but not if I give it the position that it needs to promote to a rook with no memory of hash tables from previous moves).
I only gave komodo the diagram that it should promote to a rook and it
suggested promotion to a queen with a wrong mate score.
pichy wrote:It is very simple to test it, simply use that same computer that you are using to to post here, and first play this ending versus Komodo 32 3 32bit as White and then reverse side you take Black and let Komodo play with White
Simple? Without having Komodo? How do you imagine that?
pichy wrote:It is very simple to test it, simply use that same computer that you are using to to post here, and first play this ending versus Komodo 32 3 32bit as White and then reverse side you take Black and let Komodo play with White
Simple? Without having Komodo? How do you imagine that?
I don't get it. If Komodo can't see the stalemate, it should always think the position is won, wheter it considers Rook promotions or not, because promotion to Q in the Saavedra line also easily wins if you discard the stalemate. Yet, in the founding post of this thread, Komodo reports 0.00 scores. The only way to get those is by thinking you can force a stalemate.
Are we talking about different Komodo versions now?
That is the point here Komodo 3 can't predict the next best move, if Komodo see the correct move following promotion to a Queen it should also play Rc4+ NOT Rd4-a4 to force the Stalemate.
PS: Notice that Komodo think like that only when you give the Black side to play, but if you give it the White side it will find the correct underpromotion to a Rook. I would like to see the Analysis of Don Dailey experimental program (to be released soon) which he already fixed.
Here is some analysis of the latest development version of Komodo from the original position. It has failed on the 27th iteration but the score is huge so I think it will announce mate, but I cannot wait for it as it is taking a long time. Score is over 18 pawns at depth 27.
I don't get it. If Komodo can't see the stalemate, it should always think the position is won, wheter it considers Rook promotions or not, because promotion to Q in the Saavedra line also easily wins if you discard the stalemate. Yet, in the founding post of this thread, Komodo reports 0.00 scores. The only way to get those is by thinking you can force a stalemate.
Are we talking about different Komodo versions now?
That is the point here Komodo 3 can't predict the next best move, if Komodo see the correct move following promotion to a Queen it should also play Rc4+ NOT Rd4-a4 to force the Stalemate.
PS: Notice that Komodo think like that only when you give the Black side to play, but if you give it the White side it will find the correct underpromotion to a Rook. I would like to see the Analysis of Don Dailey experimental program (to be released soon) which he already fixed.
Here is some analysis of the latest development version of Komodo from the original position. It has failed on the 27th iteration but the score is huge so I think it will announce mate, but I cannot wait for it as it is taking a long time. Score is over 18 pawns at depth 27.
I found that when I play against it komodo3 can promote to a rook and does not miss the stalemate earlier in the analysis that does not consider the rook promotion.
I did not play till the end and stopped to play after komodo promoted to a rook and here is the game
I also found that without playing the previous moves komodo3 promotes to a queen with a wrong mate announcement.
Uri Blass wrote:I found that when I play against it komodo3 can promote to a rook and does not miss the stalemate earlier in the analysis that does not consider the rook promotion.
I did not play till the end and stopped to play after komodo promoted to a rook and here is the game
I also found that without playing the previous moves komodo3 promotes to a queen with a wrong mate announcement.
Uri Blass wrote:I found that when I play against it komodo3 can promote to a rook and does not miss the stalemate earlier in the analysis that does not consider the rook promotion.
I did not play till the end and stopped to play after komodo promoted to a rook and here is the game
I also found that without playing the previous moves komodo3 promotes to a queen with a wrong mate announcement.
Now try playing with the White piece and Komodo with the Black and you will see that Komodo does not predict you to promote to a Rook
Now if you play with White and Komodo with Black it can not predict that you will promote to a Rook and it set up the trap thinking that you will promote to a Queen in order to respond Rc4+ and force mate.
And here is the famous Saavedra ending study game that I played versus Komodo:
I don't get it. If Komodo can't see the stalemate, it should always think the position is won, wheter it considers Rook promotions or not, because promotion to Q in the Saavedra line also easily wins if you discard the stalemate. Yet, in the founding post of this thread, Komodo reports 0.00 scores. The only way to get those is by thinking you can force a stalemate.
Are we talking about different Komodo versions now?
That is the point here Komodo 3 can't predict the next best move, if Komodo see the correct move following promotion to a Queen it should also play Rc4+ NOT Rd4-a4 to force the Stalemate.
PS: Notice that Komodo think like that only when you give the Black side to play, but if you give it the White side it will find the correct underpromotion to a Rook. I would like to see the Analysis of Don Dailey experimental program (to be released soon) which he already fixed.
Here is some analysis of the latest development version of Komodo from the original position. It has failed on the 27th iteration but the score is huge so I think it will announce mate, but I cannot wait for it as it is taking a long time. Score is over 18 pawns at depth 27.
If Komodo is playing Black the second move is NOT an Improvement over Saavedra Endgame study, in that case it allows White to immediately get a Queen without any chance of setting up the Known trap with Rc4+
PS: After this position Black allows White to get an Immediate advantage, you don't want this if Komodo is Playing with the Black pieces:
Well, that is what Marco has been explaining to you the whole time: setting up the trap is a blunder, DTM-wise, because it gets you checkmated very quickly (if the opponent does not let himself be suckered into it). Komodo of course sees this, so it prefers to avoid the futile trap, and play a better move in stead.
