Author: Poschmann
Date: 03:15:54 08/01/00
Go up one level in this thread
On August 01, 2000 at 03:00:03, blass uri wrote: >On August 01, 2000 at 02:48:18, Poschmann wrote: > >>On July 31, 2000 at 02:25:37, Poschmann wrote: >> >>>In a magazine I found the following position. It should be a draw ? >>>Unfortunately I don't have the following issue of the magazine with the >>>solution. >>>It is a 6-piece-endgame. I tried to solve it with the help of Fritz6 >>>(without tablebases), but I can't. >>> >>>[D]K7/8/4k3/1R1R4/8/8/p7/7r w - - >>> >>>I had the following to ideas: >>>a) Go to a 5-piece-endgame: >>>1. Re5+ Kf6 2. Rf5+ Kg6 3. Rg5+ Kh6 4. Rh5+ Rxh5 5. Rb6+ Kg5 6. Ra6 Rh2 >>> >>>b) Allow Black to promote the pawn: >>>... 4. Rge5 a1=Q+ 5. Ra5 >>> >>>I think in both cases White loses the game. But I can't proof it without >>>tablebases. >>>- Can somebody with the 5-piece-tablebases proof position a) >>>- Are some 6-piece-tablebases available to proof position b) without a pawn or >>> the initial position with one pawn. >>>- Is there somebody who knows the initial position and its solution ? >>> >>>Thank you >>>Ralf Poschmann >> >>Thank you all ! >>At first I was very surprised about the very different lines of the computer >>(crafty + tablebases) and the human. But then it become quite clear. If both >>lines come to a draw, the computer selects its best. It depends on the number >>of plies until the draw or if equal on the sorting order of its moves or >>tablebase entries in memory. >>But a human player thinks completely different. _NO_ tournament player will make >>1. ... Kd6 because this draw is seen immediately. _EVERY_ black player will try >>to win the position with Kf6 because the position is not so clear. >>If the computer should play more human like, the programmer needs an algorithmic >>expression for the different situations after the two moves Kd6 / Kf6. This >>expression will increase the practical difficulties of his human counterpart. >>Ralf Poschmann > >The next version of crafty will not play moves like Kd6. > >I suggested Bob Hyatt to evaluate tablebases draws as +0.01 for the side with >the material advantage and not as 0.00 > >Bob Hyatt replied that it is a good idea and that this idea will be used in the >next version of Crafty. > >Uri You are probably right. The first stable calculation of Crafty (I have seen it in another thread) is in ply 14. 6.Qxb6 is ply 12, that means within the horizon. From Qxb6 on there are only 5 pieces on the board. If the value is +0.01 the program tries to reach this position instead of the stupid KR-KR or the draw by repetition. Ralf
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.