Author: James T. Walker
Date: 15:16:25 04/09/02
Go up one level in this thread
On April 09, 2002 at 17:22:49, Dann Corbit wrote: >On April 09, 2002 at 17:14:55, Ricardo Gibert wrote: >[snip] >>This explains "the rating drift" perfectly. >> >>The first few years, the ratings remains about the same, then suddenly about 100 >>rating points disappear in the year 2000. Most of DC's examples follow the same >>pattern. >> >>I think DC will be disappointed that his "discovery" is not going to exactly >>rank right up there with the theory of continental drift ;) > >Actually, it is very nice to understand why. However, it is very puzzling that >the ratings do not move at all then, since much stronger programs are introduced >over time. My only explanation can be that the ratings are not recalibrated for >the whole set. The older ratings must be "cast in stone" somehow. Well look at the stats for the machine you posted. Since 1996 it only played 66 games with a total of over 1200. That will not change the rating much. Also did you see my post on "Book Learning"? My test of Fritz 7 vs Fritz 7 with 1700 games of book learning shows it has literally no effect on Elo rating. It's main function seems to be simply avoiding repeated losses. This is also understandable in my opinion since most commercial programs have thousands of variations to choose from. Rebel has over 41,000 lines. So for 1700 games of book learning the effect seems to be negligeble. Jim
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.