Author: David Rasmussen
Date: 11:37:22 06/11/01
Go up one level in this thread
On June 11, 2001 at 14:08:52, Gian-Carlo Pascutto wrote: > >Uhh something is wrong here. Using Peters utility, which, as far as I >can tell, provides correct results and takes the draws correctly into >account, I confirmed that the numbers in the table are correct as long >as all games are decisive. There is no "correct" results. It all depends on how you model the problem. The simplest would be to assume that the probability of A beating B is constant p. Then you'll get a polynomial distribution. But this is just one model, and not a very good one, for real-life purposes. In the case of humans, p is not constant, but varies with psychological, physiologocal and other factors. Also, if this bernoulli model with a constant p for one entity was true, chess ratings would be transitive. But they aren't. A can beat B, who can beat C, who can beat A, even in the case of computers (or maybe especially in the case of computers). A model closer to reality would model this behavior more closely.
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.