Author: Uri Blass
Date: 12:50:01 02/18/03
Go up one level in this thread
On February 18, 2003 at 15:26:42, Charles Worthington wrote: >On February 17, 2003 at 18:56:38, Dann Corbit wrote: > >>On February 17, 2003 at 12:02:37, Joshua Lee wrote: >> >>>On February 16, 2003 at 05:39:09, Rajen Gupta wrote: >>> >>>>rather than call these scores"ELO" lets take them for what they really are >>>>"SSDF" points; and all they measure is the relative performance of a computer vs >>>>another. this way even if it reaches 3000 in a couple of years we will not tear >>>>out hair over whether they are actually 3000 ELO or not >>>> >>>>On February 15, 2003 at 19:34:26, Jorge Pichard wrote: >>>> >>>>>If you consider that it took Deep Junior a Quad 1.9 GHz or 8 x 1.6 GHz to get a >>>>>draw score against Kasparov, then it is reasonable to suspect that Shredder 7 >>>>>can NOT possibly get a SSDF of 2768 on a mere 1.2 GHz. Either the SSDF adjust >>>>>its rating by droping another 50 points or it is time to get an Athlon XP 2400+ >>>>>to justify its inflated rating system. >>>>> >>> >>> >>>I've been saying this for years .... now someone else agree's about time >>>thankyou. >>> >>>You can also see the same if you look at Deep Junior from Dortmund and See the >>>same discrepancy. >> >>Anyone who imagines that SSDF ratings are identical to FIDE ratings is >>hallucinating. >> >>Surely, there is some sort of correlation (e.g. higher FIDE ratings will do >>better against higher SSDF ratings) but exactly what the correlation is cannot >>be said for certain. >> >>The current match, consisting of a handful of games, does nothing at all to >>correlate SSDF and FIDE ratings, since neither hardware not program from the >>SSDF list has been used. Furthermore, autoplayer must be employed, if it was >>used in the list (there are some exceptions, like CM). If SSDF hardware and >>software were used, another couple hundred games would be needed to tighten the >>reigns on any possible strength connection. > >Actually the SDDF system seems to be a joke as is their testing methods. First >of all the "deep" versions of these programs were designed to run on two to >eight threads to play at optimum strength. No The Deep fritz versions were designed to be better than previous versions in one processor. people who use more than one processor are small minority. I bought Deep Fritz6 and not Fritz6 because Deep Fritz6 was better for one processor. Not one. When you deprive the deep >programs of the extra threads you severely cripple their performance. Maybe by >60 elo or better. If a company puts a deep program on the market then they >intend to sell that program primarily to those with dual or better systems. No Chessbase claimed that Deep Fritz is better on one processor and they were right. They intended to sell it for everyone who want it. The >sddf needs to start testing the (and rating) the deep programs in multithread >mode as they were intended to be used and purchased by the consumers. This will >give the consumers a more accurate picture of the deep programs playing strength >vs. the single cpu engines. Do you have data that say that the majority of the consumers of Deep Fritz have more than one processor? Now if this system of rating tends to make the >single cpu engines look bad then so be it. The truth of the matter is that they >ARE bad. The truth of the matter is that most customers do not care if they are bad by your definition because they have only one processor. One visit to the chessbase GUI will have fritz7 and fritz 8 owners >scrambling to the web pricing dual systems. The sddf needs to wake up and start >testing these engines in the context that they were meant to be purchased in and >not the 1.2 GHZ fantasy world where all machines are equal. :-)) No The ssdf cannot test on the hardware that the richest people have. It is better if the ssdf test for the users and a lot of people even have not 1.2 GH I use only AMD 1 Ghz. 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.