Author: Ed Schröder
Date: 01:06:40 01/10/98
Go up one level in this thread
>Posted by Howard Exner on January 09, 1998 at 21:03:51: >>r3rk2/ppq2pbQ/2p1b1p1/4p1B1/2P3P1/3P1B2/P3PPK1/1R5R w - - id"Seiriwan - >>Sokolov,A" bm Qxg7+; >I am always interested in the problems that program X finds quickly >while program Y struggles. It makes me curious as to what makes them >tick. Rebel 8 for example finds WAC 141 rather quickly but has a very >hard time finding this above mate in 7. I think it is the quiet move, >Bf6 that is the stumbling block for Rebel. I checked with Rebel9. Rebel9 needs 22:00 and 11 plies. The "killer" here is not 3.Bf6 but 6.Rh1 The main variation found: 1.Qxg7 Kxg7 2.Rh7+ Kxh7 3.Bf6 g5 4.Be4+ Bf5 5.Bxf5+ Kg8 6.Rh1 and mate on the next move. I am currently working on a better tactical version. It finds 1.Qxg7 in 3:28 on ply 9 but the version in question is 25-30% slower in normal positional play. As always the big question remains if "improved" tactics will result in an overall improvement. For the moment I have baptized this version as Rebel Turtle and I hope auto232 will tell me more. I receive a lot of input from customers. Sometimes they tell you something you will not forget your whole life. One of them once said: "You guys release new versions and every new version only does better on points where it was already good" While I believe this is not the entire truth the man has a point :)) Or? - Ed - >I'm not a programmer but follow with interest this recent Singular >extension thread. It sounds like a similar technique humans employ >of giving themselves two moves in a row to help determine tactical >candidate moves.
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.