Author: Uri Blass
Date: 02:34:58 11/23/00
Go up one level in this thread
On November 23, 2000 at 04:57:53, Ricardo Gibert wrote: >On November 23, 2000 at 03:15:48, Jouni Uski wrote: > >>[D]5k2/8/pppppppp/8/PPPPPPPP/8/8/5K2 w - - >> >>Tim Krabbe in chess curiosities claims, that d5! is winning. He gives >>d5 exd5 >>exd5 cxd5 >>a5 bxa5 >>b5 axb5 >>cxb5 Ke7 >>b6... >> >>Is this true? For absolute truth we need 18 piece TBs, so it takes a lot time... >> >>Jouni > >Yes. To help get a handle on the position, it is easier to understand the >position with the e-pawns removed. In that position, the 4 vs 4 pawn >constellation must generate a passed pawn as far on the far side of the board as >possible. The rule of thumb to accomplish that is to start the action on the >near side first, hence the move d5. The presence of the e-pawns does not >significantly affect the play. I annotated this ending in detail here at CCC >about a year ago. I did not see it but I invented a smiliar position when all the white pawns were on the 5th rank and the black pawns were on the 7th rank one or two years ago and was surprised by the fact that my programs cannot solve the position. I post the position from my memory I did not use my chessbase engine to post this diagram and I hope that there are no mistakes(It should be white Ke4 pawns a5-h5 black king e8 pawns a7-h7) [D]5k2/pppppppp/8/PPPPPPPP/5K2/8/8/8 w - - I think that this is a good test position for chess program. How much time do they need to solve this position? I do not give the first move because there is probably more than one winning move(I checked only one move to be sure about it) and the problem is not to find the right move but to find it for the right reason(a jump in the evaluation by at least 1 pawn suggest that the program can see the win for the right reason). I remember that Genius3 wanted to play Kd4 in this position that is illogical move(black can play Kd8 with repetition of the structure of the position so it is clear that Kd4 is not the fastest way to win the game). I also suspect that black can defend after Kd4 by b6(I did not do a lot of analysis but I remember that I found that I can draw against Genius3 after Kd4 b6) > This position is relatively easy for a human, but hopeless for >a computer. I do not believe that it is hopeless for a computer. I believe that a computer with the right knowledge about unstoppable passed pawns based on the square of the passed pawns and the square of the opponent king can solve the position. I do not know if there are programs with this knowledge but even if there are no programs with this knowledge it does not prove that there will never be programs with this knowledge. I also believe that if you give programs enough times they are going to solve the position even without more knowledge by search. The question is only if it is going to take hours,days ,weeks,monthes or years. I guess that there is more than one solution for white to win the game(I did not analyze in order to check it) so by solving I do not mean to find the right move but to see big evaluation of at least +2. Uri
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.