Author: Sune Fischer
Date: 14:30:08 07/15/05
Go up one level in this thread
On July 15, 2005 at 14:58:32, Tord Romstad wrote: >On July 15, 2005 at 14:39:53, Steve Glanzfeld wrote: > >>On July 15, 2005 at 12:54:25, Dann Corbit wrote: >> >>>>And therefore, the games are vastly inferior in quality (as well as quantity). >> >>Vastly inferior? SSDF is testing 40/2h with 1200 MHz. CEGT is testing 40/40m >>with 2000 MHz. That is not much different in terms of CPU cycles per move. >> >>1200 MHz*180=216 billion cpu cycles per move >>2000 MHz* 60=120 billion cpu cycles per move > >No. You forget about pondering. CEGT games are played on one computer, >without pondering. This means that the actual difference is bigger than 2:1. What is quality but our subjective, often feeble minded, opinions. Two hour something games is "quality" but one hour is "garbage"?? Come on that is not rational thinking. How many people actually go over these tons and tons of automated games anyway.. In order to construct a usable and interesting rating list priority number 1 is to have enough games for a reliable rating, otherwise it _is_ going to be statistical garbage. My personal intuition is that it requires about 1000 games per engine to have fairly accurate estimate, 2000 games is better. The timecontrol should be adjusted accordingly so you can reach at least that many games before publishing the rating list. >>And of course, within the same time CEGT can provide a larger quantity >>of games than SSDF. > >Yes. That is one of the many reasons why we are very lucky to have >both lists. The two lists serve entirely different purposes, and one >of them cannot possibly replace the other. > >The time controls, pondering and hardware are not the only differences. >There is also the matter of opening books. The SSDF list is still the >only one which tests under strict tournament conditions. Each engine >has the whole computer for itself, slow time controls, own book. As >you point out, this has the obvious disadvantage that it takes a very >long time to obtain reliable ratings, and that it is never possible to >test most weak engines like mine. Therefore the CEGT list is a very >valuable complement. The rating lists are different indeed and it is good to have them both, but I tend to look at CEGT more than SSDF now. SSDF is overdoing it a little bit, time could be better spent. The fact that they have 20 fritzes doesn't exactly add to the excitement IMO. -S. >Tord
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.