Author: Peter Berger
Date: 10:19:12 07/05/01
Go up one level in this thread
On July 04, 2001 at 22:29:56, Robert Hyatt wrote: >>>But in a human match they definitely aren't. >> >>pete >> > > >I don't think they are independent events with comp vs comp either. Book >learning. Position learning. Both "connect" a series of games together. >Even more so if one program has those and the other does not. > > This is the crucial question in fact , isn't it ? If one has learning and the other one hasn't there should be a strong and bad effect on the calculations . We can safely assume that all the Chessbase programs use the same book learner ( as learning is done by the GUI ) . Let's forget about position learning ( its effect doesn't seem to matter very much and what I have seen looks quite primitive still IMHO and shouldn't matter often ). But the question is the book learner . Let's assume all programs use the Crafty book learner . Will the effects of the "learning" of the two programs neutralize each other ( which should be similar or near to independence of the games ) or will both programs' learning lead to an even stronger dependency between the games ? Or something else ( kind of a lateral effect ;)) ? I don't see that one of the possibilities is evident . I _believe_ the first alternative is true but I can't even give very good reasons for it . If the effect adds up Christoph Fieberg's program can't be used to calculate the likelihood of a match and the statements made about the Junior-Fritz match are irrelevant. Also it is not clear how your booklearner could be added to Christophs model ( which otherwise sounds rather perfect to me ) . Any ideas ? pete
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.