Chessqueen wrote: ↑Sat May 01, 2021 4:48 pm
I do NOT have your excellent Stash v30.0 to test this but with an extra White Bishop, I wonder how Stash and Cheng 4.40 do with this position, White to play and Win with NO 7 men EGTB or 7men Syzygys
[d]2b4k/8/5Pr1/5N2/8/8/8/K1B5 w - - 0 1
Cheng 4.41 would play f7 and win, troll
8 threads, 4G hash f7 +76.11 D32
EDIT: if you had any clue what you're talking about then you'd pick a 7-man position that's a draw but only with perfect play and then Cheng might (would) actually lose that
Chessqueen wrote: ↑Sat May 01, 2021 4:48 pm
I do NOT have your excellent Stash v30.0 to test this but with an extra White Bishop, I wonder how Stash and Cheng 4.40 do with this position, White to play and Win with NO 7 men EGTB or 7men Syzygys
[d]2b4k/8/5Pr1/5N2/8/8/8/K1B5 w - - 0 1
Cheng 4.41 would play f7 and win, troll
8 threads, 4G hash f7 +76.11 D32
EDIT: if you had any clue what you're talking about then you'd pick a 7-man position that's a draw but only with perfect play and then Cheng might (would) actually lose that
The obvious f7 is NOT an automatic win " Houston, we've had a problem here.”, specially after Rg8!, Here Cheng 4.40 is having a problem solving it, Probably Cheng 4.41 might do better. Now you made a fool out of yourself, since you seems to have NO clue of the difficulty of adding an extra Bishop and placing the Rook where I placed it. I know that I added a piece making it 7 men, that is to see if strong engines reach a limit with 7 men pieces
Chessqueen wrote: ↑Sat May 01, 2021 5:19 pmThe obvious f7 is NOT an automatic win " Houston, we've had a problem here.”, specially after Rg8!
this position is about all knbk, which is a knowledge I have (but far from perfect) - Cheng 4.40 can beat SF8 (the only SF I have here) in this position
CT800 (Rasmus) plays knbk perfectly though.
So I claim CT800 would win this position with white against any SF (pick one) with 6 or whatever man TBs, let's say 2 sec/move and 32M hash, whatever
feel free to prove me wrong
EDIT: ok seems I was wrong after all. CT800 1.42 failed to checkmate SF8 in knbk in that position...
Here are two drawn positions that Seer v1.2.1 failed in during TCEC S20 likely in large part due to its lack EGTB support. Seer v2.0.1, fortunately, has no such endgame issues. Many strong engines make similar errors without EGTB in my testing, though Stockfish and Komodo might have sufficient endgame knowledge.
Chessqueen wrote: ↑Sat May 01, 2021 5:19 pm
The obvious f7 is NOT an automatic win " Houston, we've had a problem here.”, specially after Rg8!, Here Cheng 4.40 is having a problem solving it, Probably Cheng 4.41 might do better. Now you made a fool out of yourself, since you seems to have NO clue of the difficulty of adding an extra Bishop and placing the Rook where I placed it. I know that I added a piece making it 7 men, that is to see if strong engines reach a limit with 7 men pieces[/size]
<snip>
Wdym "f7 is not an automatic win" ? It leads to a perfectly winnable KBNK. Stash goes for this one too:
connor_mcmonigle wrote: ↑Sat May 01, 2021 7:17 pm
Here are two drawn positions that Seer v1.2.1 failed in during TCEC S20 likely in large part due to its lack EGTB support. Seer v2.0.1, fortunately, has no such endgame issues. Many strong engines make similar errors without EGTB in my testing, though Stockfish and Komodo might have sufficient endgame knowledge.
[d]8/p7/1b2r3/3k4/2N2K2/8/8/R7 w - - 16 60
c4b6 loses.
[d]2k5/1p6/2p1K3/6P1/6n1/P3P3/8/8 w - - 1 53
g5g6 loses and only e3e4 maintains the draw.
Stash fails to understand why Nxb6 loses on the first one for several minutes, but very quickly solves the second one:
It might be due to the fact that I have no eval specification for KRPvKR. These positions are occuring quite often in games, I might take a loot at it for v31.
Chessqueen wrote: ↑Fri Apr 30, 2021 3:04 pm
I noticed that most of the testing are done with either 5men and 6men Syzygys, but have you noticed any game where the top engines has lost a single game if it does NOT use 6men Syzygys ? The challenge is if you can provide the fen where either Komodo Dragon or Stockfish 13 NNE can NOT find the correct sequence of moves without using 6men Syzygys ?
[d]3r2r1/8/k7/8/8/8/3K4/1Q4N1 w - - 0 1
White to Play and mate in 174, looks difficult for engines without EGTB support. will probably just draw this game.
Maybe even more difficult for 7-men EGTB positions. In my opinion EGTB should be handled by the GUI not the engine.
I told my wife that a husband is like a fine wine; he gets better with age. The next day, she locked me in the cellar.
Chessqueen wrote: ↑Sat May 01, 2021 5:19 pm
The obvious f7 is NOT an automatic win " Houston, we've had a problem here.”, specially after Rg8!, Here Cheng 4.40 is having a problem solving it, Probably Cheng 4.41 might do better. Now you made a fool out of yourself, since you seems to have NO clue of the difficulty of adding an extra Bishop and placing the Rook where I placed it. I know that I added a piece making it 7 men, that is to see if strong engines reach a limit with 7 men pieces[/size]
<snip>
Wdym "f7 is not an automatic win" ? It leads to a perfectly winnable KBNK. Stash goes for this one too:
Your engine Stash v30.0 is very good at finding Mate without using EGTB, the reason why I said that it is not an automatic Win, is because after f7 there is a lot of calculation that any engine has to go through in order to eventually find the Mate . Again congratulations for your excellent engine which is progressing very rapidly to be in the very top. I believe with best play from both side Black Diamond found +M53, Mate in 53.
This is where I found this puzzle, that was composed in 1951. The best move after f7 according to the composer is Rook a6+, but I believe that Rg8 which was found by your engine and Black Diamond prolong the mating twice as long ==>
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
That is correct if anybody wants to find out if any 5 men or 6 men ending is a draw or a win all they have to do is to use an endgame Table, but we are trying to find out what engines perform the best in finding either a draw or a win without using any 5men or 6men EGTB. As long as an engine support any type of endgame table it does not matter what rating it has it could be as low as 1400, but with a 5 men or 6 men EGTB, simply feed the fen and it will not missed a draw or win. I just noticed that certain engines have more endgame knowledge than other for instance Stash v30.0 can usually find a particular 5 men or 6 men ending faster than Cheng 4.40., therefore, even if they are close in rating if they play a 1000 games without using any EGTB, Stash will score more ending win and save more draws than Cheng.
NOTE: Can you imagine if Carlsen play against 4 GMs rated around 2550 thru 2650 but they could have access to any opening book of up to 20 moves and 7 men EGTB using a computer, while Carlsen has to rely on his knowledge, all they have to do is survive the middlegame to beat Carlsen.
