Author: Uri Blass
Date: 08:52:34 11/10/01
Go up one level in this thread
On November 10, 2001 at 10:50:25, Thomas Mayer wrote: Hi Thomas, >Hi Uri, > >>> oh my god, Uri, I really do not want to calculate any rating out of 9 games. >> It would be unfair to ChessTiger, think about the hardware difference. > >> I disagree. >> chess rating is something fair. >> I do not believe that your program is going to get higer ssdf rating if it >> plays only against programs with slower hardware. >> It is going to win more games but it does not mean that the rating is going to >> be bigger. > >So why does Christoph not allow me to test against the top programs on the SSDF? >At least ChessTiger on Palm should get some points against them to prove that it >is at 2100... I doubt that it would get any points against Gandalf, Junior 7, >Fritz 6, Shredder 5, the Tigers when they are playing on Athlon 1200... The difference between 2100 and 2600 ssdf rating means that it is going to get almost nothing and testing when the rating difference is too big says almost nothing. The problem is about the difference in rating and not about the difference in hardware. I am sure christophe has no problem with testing palm tiger against ssdf programs of similiar rating and faster hardware(there are p90 programs with similiar rating(I think cometA32 is one of them). There is no point in match with rating difference of more than 400 elo and it is better to avoid also matches with rating difference of more than 200 elo. And >anyway I am quite sure that it has at least 2100, maybe even more... >He says himself that the difference of speed is to big here... He means that the difference in speed is too big to get similiar rating. > >>> It is based on winboard engines. What is so wrong with that ? No need to >>> include any commercials, crosscomparisson is made with Crafty, SOS & Little >>> Goliath... > >> It means that the rating is not comparable with the ssdf rating so the rating >> of the winboard is meaningless to get an estimate for the rating against >> humans. > >So, following your argumentation, the SSDF-list is also meaningless to get an >estimate for the rating against humans. Not exactly The ssdf list is based also on games against humans. The estimate based on the ssdf may be wrong but it is at least based on some information of games against humans when the winboard list is not based on information of games against humans. There is not more crosscomparisson then >the winboardlist has with Crafty, SOS & Little Goliath. Agreed ? > >> I do not believe it. >> I think that you underestimate Quark. >> I guess that quark is going to get similiar rating in both cases and I will be >> surprised if it is going to get below 2200. > >most programmers underestimate their own creation... :) But I am quite sure that >it would not get much more then 2200 maybe a lower rating then that on the SSDF >list if only playing against the tops also on Athlon 1200... > >Greets, Thomas I believe that you are wrong. If it can do only 25% against the top programs then it means that it only 200 elo weaker then them. I also know that top programs often have problems to beat weaker programs on slower hardware. I remember that shredder2 started well against many of the newer programs on better hardware in the ssdf games. Shredder2 had no chance to win matches because it had no learning function but there were even cases when it was leadng after the first 6 games. I remember that Junior5(p200) lost rating from playing against old Comet(CometA32 if I remember correctly) on p90. It could only win 17.5-2.5 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.