Author: Drexel,Michael
Date: 23:11:49 01/29/04
Go up one level in this thread
On January 29, 2004 at 17:06:44, Stefan Zipproth wrote: >On January 29, 2004 at 15:51:58, Drexel,Michael wrote: > >>On January 29, 2004 at 12:20:35, Stefan Zipproth wrote: >> >>>Hi all, >>> >>>thanks for your comments. Because of the big interest in this topic and somewhat >>>unfriendly e-mails that were sent to Fritz Reul, we decided to put an official >>>statement on the web site (see "About List"). >>> >>>I want to make clear that we do not want to argue and raise hot discussions, or >>>achieve that everybody agrees with our statement. But we definitely want to >>>achieve that people talk about chess again - even if this is a topic that is of >>>course important and which gives reason for discussions. >>> >>>Best regards, >>>Stefan >> >>You wrote: "It is easy to prove that List is far stronger than any version of >>Crafty" >> >>Can you elaborate on that? Have you tested at least one of the recent versions >>19.09 winboard or 19.08 CB? >>Do you refer to tournament time controls? >> >>Michael > >Hi Michael, > >thanks for your remakrs. I never refer to tournament time controls, because I >see no way to play hundreds of games with long time controls. If Crafty would >win 13-7 against List, this is has no relevance because it is not statistically >significant. All depends on your testing method. Crafty scores probably almost 50% against List 5.12 if you use the large Fritz8 opening book for both engines. Crafty might be (or it is most probably) weaker in the middlegame, but it can outplay List in many endgames convincingly. List has many strategical weaknesses. Of course you would have to use a recent executable (Crafty 19.09,19.10 not the latest CB engine) and make it aware it's playing a computer. Just to match two engines is not sufficient anyway. The problem is: If you would test them against other engines it also depends. You can always find a subset of engines where Crafty scores better than List on average. > >My tournaments often begin with something like 13-7, then it continues with >30-40, and only after several hundred games the result becomes stable. Simple >mathematics, as mentioned in section "Ratings lists". Thanks, I know about the basics of probability and statistics as well. > >In addition, I must admit that I did not test against Crafty, but a long test >against Fritz 7 was made, with time control 5+5 if I remember correctly. List is >as strong as Fritz 7. I do not know if Crafty does indeed achieve this. > >Besides: Long games do not increase the significance, i. e. 2 long games are no >substitute for 4 blitz games - they are only a different way of testing which >produced different results. But as I said, I see no way to use these test >conditions (though I would like to), only by analyzing long games by hand. But >that is only useful for improving an engine, because one cannot conclude from >one or 10 games to the exact playing strength of an engine (or a human player) - >chess is too multifaceted. You wrote "it is far stronger". That's not the case. Even 100 ELO better is not far stronger. I am far stronger than chessplayers with national ratings about 1850 DWZ. For sure I am not far stronger than someone with 2150 DWZ. (Both ratings based on > 500 tournament games for example) Michael > >Best regards, >Stefan
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.