Author: José de Jesús García Ruvalcaba
Date: 13:47:23 02/24/00
Go up one level in this thread
On February 24, 2000 at 15:31:55, Dann Corbit wrote: >On February 24, 2000 at 15:15:54, blass uri wrote: >[snip] >>If the ssdf rating of shredder2 should be zero then there is a big problem >>because the rating is based on the assumption that shredder2 has rating close to >>2500. > >After an infinite number of games against programs that learn, the ELO would be >zero Why zero? According to my calculations, if a program loses all but a finite number of an infinite sample, its rating should be minus infinity. José. >(assuming that the learning is not full of bugs). However, from a >practical standpoint it does not matter, since we won't ever play an infinite >number of games. > >On the other hand, games that do not learn should be "punished" since they will >get beaten by learners in real-life, given enough time.
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.