Author: Sune Fischer
Date: 13:05:09 07/12/03
Go up one level in this thread
>> >>> >>>I certainly believe that there are indications that the win/lose probablity >>>pridicted by the Elo scale is scewed and does not correlate with practice very >>>well when differences are large. >>> >>>-S. > >Do you think that the weaker program is going to earn rating or lose rating when >the difference is more than 300 elo? > >Uri I remember one of the graphs he had done, was a plot of rating difference versus the winning probability for the highest rated of the two players. Where the Elo formula pridicted one probablity curve through the graph, actuality showed that this curve was quite a bit lower, i.e. that higher rated players didn't on average manage to score as high as pridicted. The bigger the Elo difference, the worse was the fit. This of course means, that the Elo model tends to shrink peoples rating when they play weaker players. Since the top guys almost constantly are playing weaker players, they are bound to suffer from this deflation effect. It shouldn't be very hard to repeat his experiment, just take a large database and look at all the games where the difference is from 195 to 205 points and see who won and how often. -S.
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.