Author: Uri Blass
Date: 01:08:45 08/27/02
Go up one level in this thread
On August 27, 2002 at 02:18:43, Uri Blass wrote: >On August 26, 2002 at 23:34:15, Will Singleton wrote: > >>It's common wisdom that a good book will increase a program's strength, and I >>have no doubt about it. But, what about the typical amateur book, which is >>usually created from pgn files, without a great deal of hand tuning? > >I do not know what is the typical amateur book. > >The book of the latest movei that is not public >is not created from pgn files but from few lines that were added >manually. > >I usually do not test it with book but with the option of always >change first move in order to avoid double games because I do not >have an easy way to avoid double games without it in a match >of 40 games. > >Movei00_76 could score slighly more than 50% against >amateur2 in my test(p800 no pondering when programs use 16 mbytes for hash >tables) but I suspect that against amateur >lines like 1.e4 a6 can be productive so it is possible that >movei00_76 with my small book could score less(I do not try movei >with my small book because I do not want double games and >double games can happen unless the match is of at most 4 games). > >I tried only 40 moves/1 minute and 40 moves/10 minutes matches >if I remember correctly and it is possible that the picture >could be different at longer time control(at least it seems to be the case >against Ant when Movei scored more than 50% against ant at >40 moves/10 minutes and lost against ant at 40 moves/50 minutes >26-14). > >Note that I used movei00_761 in the second match so it is also possible that >00_761 is weaker inspite of the fact that it scored in average >slightly better in the GCP test suite(there is no change in the evaluation >from 00_76 to 00_761 and the only changes that I did are supposed >to make the search more stable. > >A fast test did not suggest a big difference between 00_76 and 00_761 >and 00_761 lost against 00_76 20.5-19.5 in 30 seconds/40 moves >match. > >Uri It seems that 00_761 is worse than 00_76 based on more tests. In the last match against 00_7a(5 minutes/40 moves) it beatd 00_7a only 21-19 when 00_76 used to beat 00_7a in results like 28-12. I guess that the rules to do the search more stable were not good. Uri
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.