Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Book learning?(a neglected subject?)

Author: Dan Kiski

Date: 09:00:26 01/01/99

Go up one level in this thread


On December 31, 1998 at 19:47:37, greg moller wrote:

>On December 31, 1998 at 14:02:13, Marc van Hal wrote:
>
>>I was forgotten to tell that the booklearning option starts to say that an
>>intire opening is bad and didnt look at all subvariation first so it should only
>>lock an subvariation and try an other line perhaps the mainline is not the best
>>after all
>
>
>Wouldn't it be great if somebody could post a definitive report on how book
>learning is really supposed to work, and somehow compare a few programs'
>learning features ?
>
>For example Mchess has both book learning and "regular" learning... now what's
>the difference, and does it really work as it's supposed to ? And where exactly
>does a program's learning aftereffect actually kick in, i.e. at which point do
>you discard a branch? In fritz5 book learning seems to gradually eliminate a lot
>of the early replies until you're left with one move ( let's say 1.c4) which
>then gets played over an over, unless you reset, which does, to an extent,
>defeat the purpose of any real learning. This seems very crude. Hopefully
>someone  can point to a more intelligent learning feature. I think it's a
>fascinating subject.
>
>
>regards, greg

It seems from reading Marty's response to my original post that he has tried to
make book learning more intelligent. But you are correct as to fritz's method
yet all seem to do this when set to tournament book, at which point it is
believed the openings chosen and moves played are the strongest.







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.