Author: Bas Hamstra
Date: 14:48:48 01/09/02
Go up one level in this thread
On January 09, 2002 at 03:50:56, Uri Blass wrote: >On January 08, 2002 at 18:47:35, Bas Hamstra wrote: > >>On January 07, 2002 at 16:56:46, Uri Blass wrote: >> >>>On January 07, 2002 at 14:53:25, Bas Hamstra wrote: >>> >>>>On January 07, 2002 at 13:44:43, Joshua Lee wrote: >>>> >>>>>you are Ill informed . As much as i think the SSDF is Inaccurate in it's >>>>>ratings it does a few things that are consistent: >>>>>1 Shows which programs are stronger than others and on what hardware >>>>> >>>>>This is the only point i need to make to totally bust your argument because >>>>>if you look at the estimated ammounts they give for each of their hardware >>>>>increases or if you take the averages from one set of hardware to another you >>>>>should see clearly that in reality there is a definite increase in most newer >>>>>versions. I don't know how the program authors do it but they do..... >>>>>look at Junior 6 and 7 on a K6-2 450. >>>> >>>>I dont see your point. I am not saying that programs don't improve. But I do say >>>>that Junior 7 does not stand a chance against Junior 6 on 2x slower hardware. >>>>Can you find 1 program where a newer version beats an older on 2x slower >>>>hardware? >>>> >>>>Bas. >>> >>>Yes >>> >>>Fritz7 on old hardware beats Fritz3 on fast hardware >>>if the hardware adavantage is 2:1 >>> >>>I am not sure about Fritz7b and Fritz6 but >>>it may be interesting to test. >>> >>>Uri >> >>How do you know? Data! Numbers, tables. Photo's :-) Are you willing to take a >>bet? I say Fritz3 is going to win. >> >> >>Bas. > >I remember that the difference between Fritz6a and Fritz5 based on the ssdf is >more than 70 elo on the same hardware. >I also know that programs do not earn more than 70 elo from being >twice faster based on the ssdf. > >Fritz6a may have a better book so in order to be careful I did not suggest to >take Fritz6a and Fritz5b but to take Fritz7a and Fritz3. > >I will be surprised if Fritz3 can win a match against Fritz7 when the hardware >of Fritz3 is exactly twice faster even if you use the same book. > >Uri There are certain effects that make ELO calculations and extrapolations very tricky IMO. - books & booklearning - old programs don't play on the fastest hardware and their rating erodes. Note that old Fritzes have been *longer* on older hardware relatively. - A beats B, B beats C, C beats A - When programs are 95% alike, smalle differences get exaggerated ELO-wise - You cannot use SSDF rating differences to predict the outcome of a 5/0 Blitz match (referring to my other post). Fritz3 might be very well be *far* better at Blitz, being a extremely fast PieceSquare program. Enough to make a prediction of a matchresult based on SSDF ELO differences pretty worthless, especially if it's a blitz match. Best regards, Bas.
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.