Author: Rolf Tueschen
Date: 02:49:01 10/09/02
Go up one level in this thread
On October 09, 2002 at 00:31:49, Dann Corbit wrote: >On October 08, 2002 at 20:19:22, Rolf Tueschen wrote: > >>On October 08, 2002 at 19:55:50, Dann Corbit wrote: >> >>>On October 08, 2002 at 19:22:55, Rolf Tueschen wrote: >>> >>>>On October 08, 2002 at 19:18:27, Omid David wrote: >>>> >>>>>On October 08, 2002 at 19:00:11, Roy Eassa wrote: >>>>> >>>>>>...maybe Kramnik is just a much stronger chess player than Deep Fritz? >>>>> >>>>>I won't dispute that. >>>> >>>>But how could that be true if DEEP FRITZ has almost the same Elo number??!! >>>>Would you say that the numbers are bogus? >>> >>>There is no way to compare ELO figures from different pools of competition and >>>with different starting seeds. >>> >>>If ever there were an apples to oranges comparision, this is it. >> >> >>Not nice! Let me be the devil's attorney. > >He's got plenty of attorneys. ;-) Thanks for the nice contribution here below, Dann! > >The way ELO works is that you compare differences between different players in >the same pool of competition. The difference is all that matters. I say that the belonging to the same pool is all that matters! And that already is the greatest sin of all time in computer chess - what the SSDF is actually doing! But nobody in CC wants to listen! I repeat, and that is a great crime!! The testing of machines / progs from different pools and different epochs is a deadly sin in the SSDF. > You can >subtract 10,000 from everyone's ELO number in one pool and all the computations >of win expectancies remain the same. > >So if Joe is 1900 ELO and Fred is 1400, then (over a broad average) we would >expect Joe to score 95% of the points. > >If (on the other hand) Joe is 2900 and Fred is 2400, then (over a broad average) >we would expect Joe to score 95% of the points. > >If (on the other hand) Joe is 900 and Fred is 400, then (over a broad average) >we would expect Joe to score 95% of the points. > >Within a pool, the only thing of importance is the differences. The absolute >number is basically without any meaning. > >2800 FIDE has a meaning. Yes, of course. BTW you are thinking that Devil's advocate means redundance? >2800 SSDF has a meaning. Of course NOT! > >The two numbers do not mean much in comparison with each other. We can >reasonably expect that a 2500 SSDF player will be a lot weaker than a 2700 SSDF >player. But that is all. We cannot say how SSDF numbers might translate to >FIDE or USCF or any other sort of collection. Mr. Dann, I hereby take you under custody until you will confess that you are guilty of using the Elo numbers from SSDF for certain programs without the worthless appendix "SSDF". Will you confess? Then please repeat it here in front of the many TV cameras and the American, Swedish and German people! So help you the Devil. Rolf Tueschen
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.