Author: James T. Walker
Date: 03:13:21 03/27/02
Go up one level in this thread
On March 26, 2002 at 16:20:59, Dann Corbit wrote: >On March 26, 2002 at 15:09:03, martin fierz wrote: > >>On March 26, 2002 at 14:42:40, Dann Corbit wrote: >> >>>On March 26, 2002 at 14:17:53, Peter Hegger wrote: >>> >>>>Hello, >>>>If Gulko beats Shredder today, the comps will still have a 5-3 score and a >>>>combined TPR of 2697. >>>>If its a draw today, 5.5 - 2.5 = a 2738 TPR for the comps. >>>>If Shredder wins today, 6-2 = a 2792 TPR for the comps for the 8 game series. >>>>Granted, its not 40/2, but even at 60 10 its still a very impressive performance >>>>by the machines. >>> >>>If Gulko had played against 5 human players would we conglomerate their ELO? >> >>if we want to calculate a TPR, yes, of course. that's how it's done... > >It seems that we are doing it upside down (IOW -- forming a TPR for a >conglomerate entity even though they played single games). The TPR for the >machines would have to be based on two games. Unless I misunderstand the >algorithms. Hello Dann, Of course you are correct but lets look at this practically. TPR's are only used for unrated players to get a "quick" performance rating in a particular match/tournament. The USCF calculates it by Rp=Rc+ (400*W-L)/N. We would have to calculate each programs TPR based on 2 games. Not much use practically. But it's fun to take another view point. The SSDF had a couple of the programs rated at around 2700. The newer programs not rated yet can be assumed to be close. So if we assume an average of say 2700 for the programs we could give GM Gulko a "TPR" based on his performance vs the average 2700 Elo machines. His performance rating would be 2700-133 or a SSDF rating of 2566. But since Gulko performed at -133 compared to the programs doesn't it follow that the programs as a group performed at +133 compared to GM Gulko? :-) Sure that's where I'm going. So now we can take Gulko's known FIDE rating of 2601 and say the programs performed at around 2734 in this match on average. Neither assumption is correct but just food for thought. Jim
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.