Author: Uri Blass
Date: 04:09:02 10/21/02
Go up one level in this thread
On October 21, 2002 at 06:22:59, Daniel Clausen wrote: >On October 21, 2002 at 05:36:23, Brian Katz wrote: > >> Here is an interesting experiment involving Hash Table settings. >>This was run on a single processor 350MHz Pent II 384 SDRAM with Nalimov 4 disc >>Endgame Turbo plus additional downloaded 5 piece endgames and Permanent Brain >>turned off. >> >> For my computer's speed the proper HASH TABLE settings for game 5/12 should be >>16 MB. I have found that 32MB has always worked well, however. >>As an experiment, since I believe Fritz 7 performs better than Deep-Fritz 7 when >>using a single processor (even though Chessbase boasts otherwise, stating that >>DF7 is stronger even when usning a single processor). >> I raised the Hash Table settings for Fritz 7 to 192 MB as opposed to >>Deep-Fritz 7 set at 32 MB Hash. >> Surely, 192 should be much too high for such a time control with my processor >>speed. Well, Fritz 7 still beat out Deep-Fritz 7 by 5.0-3.0 >> >>Blitz:5'+12" 0 >> >> 12345678 >>1 Fritz 7 110½011½ 5.0/8 >>2 Deep Fritz 7 001½100½ 3.0/8 >> >> Can anyone please explain this please? > >*sigh* > >You don't realize that 8 games are a ridiculous number of games for such an >experiment? Why don't people understand that!? It doesn't take a degree in >statistics to see that! Throw a coin 8 times, and the result 5-3 won't be that >uncommon... > >Sargon Note that the "coin" that was thrown has 3 sides and not 2 sides. Another note is that 2 different "coins" were thrown 4 times: one "coin" is Fritz7 white and one "coin" is Fritz7 black. Both things should slightly reduce the probability for 5-3. I agree that 5-3 is still not a significant result but it seems that the poster already had the impression that Fritz7 is better before the match so it seems that he is based on more games. 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.