Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Forward with Opening Theory

Author: Richard Pijl

Date: 09:52:08 05/17/05

Go up one level in this thread



>This would be an interesting idea, however if you take away the opening books
>you take away any possibility of randomness by the engine (unless the engine had
>some randomness setting and the personality being used in the tournament was
>using it to some degree -- and this, of course, generally results in lower
>quality play).

This is not necessarily true, as you can see on ICC with the 'ctd' account. This
engine has no opening book, and virtually no knowledge. But to prevent human
players from repeating losing lines, it does have a simple form of position
learning (basically storing rootnode search result and depth in a permanent
hashtable and using it inside search). It gradually improves its own opening
play by trial and error. As it is pretty weak, it is quite popular among the
stronger (untitled) human players and it gets to play a lot of games.

So, although randomness is removed, it usually doesn't play the same lines all
the way.

Richard.



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.