Author: Robert Hyatt
Date: 18:12:01 04/18/02
Go up one level in this thread
On April 17, 2002 at 14:39:22, Vincent Diepeveen wrote: >On April 17, 2002 at 13:45:33, Uri Blass wrote: > >DIEP soon plays Kh1 there, if it doesn't then within >a few minutes i get a fail low to nearly draw score >for Kf1? > >>On April 17, 2002 at 13:26:34, K. Burcham wrote: >> >>> >>> >>>Deep Blue vs Kasparov 1997 Six game match >>> >>>if you have any Deep Blue positions that were played in the six game match, that >>>you feel are very difficult for todays programs, could you please post them >>>here. I will check to make sure I have them. >>>If you have any positions you feel that cannot be found by todays programs, >>>please state that with the position please. >>>I am trying to compile the complete list of Deep Blue moves that anyone thinks >>>cannot be found, or played by todays programs. probably have most--but you might >>>have one i overlooked. >>>If you have read somewhere about a Deep Blue position that someone thought was >>>too difficult for todays programs, please post with position and game number. >>> >>>question everything, >>>kburcham >> >>You should look also for deeper blue tactical mistakes that part of the top >>program understand that they are big tactical mistakes after a long search. >> >>One candidate is Kh1 from game 2. >> >>It may be interesting if people can compare the score of top programs after Kf1 >>and the score after Kh1 if you give programs 24 hours to search. >> >>Note that in order to find that Kf1 is a tactical mistake it is enough to find a >>big difference between Kf1 and Kh1. >> >>Programs do not need to find a draw score after Kf1 but only that Kh1 is >>significantly better than Kf1. >> >>Uri The programs need to see two things: 1. white is winning. 2. black, therefore, will be trying to force a draw. If, at any point in the analysis, the programs think that white is worse, they will think white will go for the draw. Which means that seeing a draw score for Kh1 vs Kf1 might not mean a thing as far as the real game goes... Seeing a draw score at the root means one of two things: 1. White wants to draw because it stands worse, and white can force a draw no matter what black does. 2. White wants to avoid the draw because it stands worse, but black can force it no matter what white does. case 2 applies to the DB vs Kasparov game. But case 1 can produce draw scores because somewhere down in the search the program becomes convinced that black stands better and therefore white will go for the draw... Forced mates are forced mates, with no such nonsense. But drawscores are very tenuous in nature and depend on how well the program understands what is going on... or not... as the case may be...
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.