Author: Roy Eassa
Date: 09:40:01 10/07/02
Go up one level in this thread
On October 07, 2002 at 11:19:25, Rex wrote: >Everyone is saying how a computer program will_not_and does not see Rd4 played >by Kramnik Ver DF in the second game. Well I got some news. The new kid on the >block Ruffian finds Rd4 in just over 10 minutes on a 1.533Ghz computer as >Ruffian also found move a4 played by Kramnik in just over 6 minutes. Folks >Ruffian needs some serious looking into. It play MORE like a top GM than a >GREAT chess program.. > The point is not finding Rd4+. Many programs see it. The point is in understanding that forcing major exchanges leads to a winning endgame for White, even though material is even. No program understands that. Ruffian's evaluation shows that it certainly doesn't. I'll bet that in Kramnik's mind, the evaluation was a lot higher than +0.19 for White -- he KNEW he would win the resulting endgame because he understood the full significance of Black's passive position. His goal was to force exchanges so he could win the endgame. Programs' goals are to play the variation that scores higher than the others. I just tested Ruffian with the position, and it considered at least 3 moves to be about equal to each other. In fact, after 3 minutes on a 1.2 GHz Athlon, it is still ranking Rd4+ third after Rd2+ and Rd1+, with all 3 evaluated between +0.06 and +0.20. That's NOT telling us that it understands that trading off pieces wins the game, much less that it would have aimed for such a target from earlier on!
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.