Author: Uri Blass
Date: 01:10:17 12/06/02
Go up one level in this thread
On December 06, 2002 at 02:37:43, Jorge Pichard wrote: >On December 05, 2002 at 18:29:31, Mike S. wrote: > >>>To require that all programs run on an "old" computer would be to require that >>>new chess engines not take advantage of the new technologies. >> >>It requires only that engines do not *need* new technologies to run at all (i.e. >>like F7 requiring MMX CPUs). But as long as they run on old computers, they can >>do that, if they take advantage of new technologies on other computers or not. >>It would just add some fun for poeple who like to argue like, "The list is >>wrong, engine XY would rank much higher if latest CPUs would have been used, 2 >>Gigbytes Rambus hash, etc.etc." :o)) >> >>(But of course it's favourable that test hardware isn't too much different from >>what's the "average" of what the fans use at home.) >> >>Regards, >>M.Scheidl > > >Yes, that is precisely the point. The main reason for establishing the SSDF was >to give the computer fans an estimated rating based on what's the average of >what most of the computer fans use at home. Two years ago the average hardwares >used was somewhere between a K6-450 Mhz and AMD 1.0 Ghz, but nowadays a more >realistic average would be somewhere between an AMD 1.2 Ghz and an AMD XP 2600+. The question is if your estimate of average is correct. I suspect that many users still use machines that are slower than AMD 1.2 Ghz and 2 years ago many users used hardware that is slower than 450 Mhz. The speed of the computers that I use today is 850 Mhz and 1000Ghz. >Of course you can always estimate what the rating would be by adding 50 Elo >points when you double the hardware's speed. But the K6-450 it is just too Old, >it doesn't matter if CT 15 performs better than Deep Fritz on a K6, when hardly >15% is still using the Old K6 450 Mhz. > >Pichard. The statistical error can tell use that we do not know if tiger15 is better than Fritz on 450 Mhz and we also do not know if Deep Fritz7 is better than Tiger15 on 1200 Mhz. Another problem is that the words better are not defined well. To be number one on 1200 Mhz you need to be better in beating weaker players and to be better on 450 Mhz you need to be better against stronger players. If Fritz is better against weaker players when Tiger is better against better players then the question which program is stronger is not defined. I think that the information about 450 Mhz gives us also important data and knowing that tiger is number one on 450 Mhz increase the chances that tiger was simply unlucky in 1200 Mhz when the statistical error played against it. 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.