Author: Uri Blass
Date: 07:49:18 11/13/00
Go up one level in this thread
On November 13, 2000 at 09:27:08, Ricardo Gibert wrote: >On November 12, 2000 at 21:17:53, Howard Exner wrote: > >>On November 12, 2000 at 18:41:13, Ricardo Gibert wrote: >> >>>Here is an endgame study that I think everybody will find entertaining: >>>[D]8/1p6/1p6/2k4P/7P/5ppK/7P/8 w - - 0 0 >>> >>>Fritz4 is hopeless in solving this position. I was wondering if newer engines >>>with 5-man EGTBs would stand a better chance of solving it. >>> >>>Hint: White Queens all three of his pawns. >> >>This is what Tiger 13 produces when given 5 seconds per side using an Athlon >>900. >> >>1.Kxg3 Kd4 2.h6 Ke3 3.h7 f2 4.h8Q f1Q 5.Qe5+ Kd2 6.Qd4+ Kc2 7.Qf2+ Qxf2+ 8.Kxf2 >>b5 9.h5 b4 10.h6 b3 11.h7 b2 12.h8Q b1Q 13.Qh7+ Kc1 14.Qxb1+ Kxb1 15.h4 1-0 >> >>Is this on the right track? > >This is all correct and it cooks Kashdan's solution to boot! He gives 6.Qf4+ >instead. See my response to Hyatt's post in this thread. > >Unfortunately, the idea was to see if programs could find the correct line from >the beginning position. You got lucky with the method you used. If it were >really a reliable way to analyze a position, chess programs would perform their >searches in a very different way than they do. Giving chess program to play against themselves is a reliable way to analyze positions if it is only one of the way that you use to analyze. chess players use this method in correspondence games. I can learn from it in cases that the evaluation say advantage for white and giving programs to play against themselves say a win for black. In these cases I try to improve the line for white. There are things that programs cannot see practically even after many hours without this idea but they can see them if you give them to play against themselves and to learn from the game. There are cases when there is a long forced line that programs cannot see because the line is too long but if you give them to play against themselves and learn they can see the long line and avoid the first move that leads to trouble. It is also possible to learn when there is no forced line. If you see that a move that is supposed to be winning is losing and playing again when you try to improve the line does not help you can learn that the evaluation function of the program is wrong and it may be better to change the evaluation(you can learn that taking the knight at g5 is probably bad and change your evaluation to avoid taking the knight). Uri
This page took 0.01 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.