Author: blass uri
Date: 05:40:25 06/01/98
Go up one level in this thread
On June 01, 1998 at 06:09:29, Moritz Berger wrote: >On June 01, 1998 at 05:31:01, blass uri wrote: > > >>SSDF did not use pentium 200MMX with 64Mbytes Memory >>to test other top programs but Pentium 200MMX with less RAM > >CAN YOU PROVE THIS? Have you any hard or soft evidence that this is >indeed the case? I read this in the past I understood that the testers needed to have at least 2 computers with 64Mbytes RAM to test other programs in equal conditions and I believe part of them did not have it I thought the reason they did an autoplayer version that can run only on computers with 64 Mbytes RAM is to have an unfair adventage. It helped fritz5 to play in better conditions and there was nothing to help other programs to play in better conditions but I apologize if fritz5 did not play in better conditions. > >>If other top programs used the maximum hash size that is inherent in >>them >>it is not important but I think it is not the case. > >Any FACTS? I read that rebel9 can use 60Mbytes of hash tables. I do not know about facts but I did not hear SSDF decided to include only games played in fair conditions. I believe the adventage is not significant(maybe 5-10 ELO points) so it is not important. > >>The unfair adventage did not help fritz5 much so probably fritz5 >>deserved to be number 1 in the ssdf List >>(it does not say fritz5 is number 1 because the >>list does not have other top programs like Junior4.6) >> >>fritz5 has another unfair adventage that the autoplayer they use >>is not public because they want to continue to be number 1 by unfair >>means >>(they can learn their opponents when the opponents cannot learn them). > >Did you know that the Fritz 5 Powerbook doesn't contain *ANY* learning >values? >Fritz even plays (and gets punished badly for doing so) the 1.c4 English >opening where it reliably loses. No book tuning at all - all the other >top programs rely on hand tuned, computer optimized opening books. Just >a few FACTS about 'unfair advantages'. > >Sorry, I don't want to start the debate all over again - it serves no >real purpose. My point is just that > >a) we don't know if the other programs got less than 64 MB RAM (this >might have been different for individual matches, for all we know, e.g. >Nimzo getting 200 MB hash tables vs. Fritz with 44 MB HT - but the point >is: we just don't know) If this is truth nimzo had an unfair adventage against fritz5 but I believe it is not a very big adventage > >b) accusing F5 of book tuning of any kind is most ridiculous when you >consider that it uses a big GM database without any learning values or >any other optimizations. It exclusively relies on aggressive book >learning to save the day. > I believe they did not use their adventage that they can learn the opponents but the point is that they can use this adventage if they see this is their only way to stay number 1. I understand that this is the reason many programs like Genius or Rebel will not be in the SSDF list Uri >Moritz
This page took 0.01 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.