Author: Serge Desmarais
Date: 21:54:10 09/01/98
Go up one level in this thread
On September 01, 1998 at 17:53:36, Robert Henry Durrett wrote: >Although there apparently have not been many games played between the top >programs and humans strong enough to win against these programs, there seems to >be a general consensus among the more computer-savvy CCC members that at least >some of the better programs are [or "probably" are] better against humans than >their comp vs comp tests would seem to suggest. True? > >For those on this CCC who know at least a little about how chess engines work, >the questions: > >(1) "What do you suppose there is about the inner workings of such programs, >which do better than expected against humans, that help these programs against >humans?" > >(2) "What is there, specificaly, about some programs which make them seem to NOT >do better against humans than their comp vs comp test results indicate?" > >If specific answers can be obtained, then what do these answers suggest for >future design guidelines for chess engines? I am still not convinced that the programs that are the top of the SSDF are not the top of computers afainst humans in normal tournament games conditions. I-Fritz 5 is top of SSDF and did very well against granmasters (true, it was lightning chess) II-Rebel is amongst the top programs at the SSDF and did very well against humans too (Anand had a hard time with it and at Aegon's last tournament, it was very impressive ; but, again, it was mostly at faster time controls than in tournament games. III-HIARCS 6.0 is #2 on the SSDF list and did beat IM Deen Hergott a 40/2 - 6 games match. I cannot think of any #20 or so program that did better than the top programs in any Aegon tournament nor in any event I can think of right now? Note: The bug reported by Bob Hyatt that did not change anything against humans but a lot against computers were in games played in ICC : mostly 5 minutes blitzes (sometimes with an increment of 2-3 seconds) and maybe a few 15 minutes games? Most GMs in ICC avoid other kind of games than blitz games. Even against a computer or program that has a grave bug, I don't feel that with 2-3 minutes on the clock and just taking less than 10 seconds to move, a GM would have had time to "refute a weak play". GMs were being beaten by computers in blitz games on much slower computers than the actual ones (even on 486 computers!)... And again, in the Aegon tournaments, it was still a bit too fast for a GM to show his full strenght (for most of them). Serge Desmarais
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.