Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is there really any good alternative to alfa-beta search?

Author: Robert Hyatt

Date: 06:49:14 06/09/98

Go up one level in this thread


On June 09, 1998 at 05:19:15, Jouni Uski wrote:

>On June 09, 1998 at 04:47:18, Ernst A. Heinz wrote:
>
>>On June 09, 1998 at 04:11:18, Jouni Uski wrote:
>>
>>>According to CSS magazine 2/98 Ulf Lorenz program P.Conners works
>>>without alfa-beta search and is still good (tactically very good).
>>>And it doesn't need any hash tables!
>>
>>"P.Conners" is based on "parallel controlled conspiracy number search"
>>(PCCNS) which needs *much* more memory than any standard alpha-beta with
>>transposition tables because it keeps the whole search tree in RAM ...
>>
>>As for tactical power, any decent program running on 40x P-II 266MHz
>>should not be too weak in this respect, should it?
>>
>>=Ernst=
>
>Thanks for info. CSS story was a little misleading and incorrect so I
>got there
>wrong info.


Also conspiracy numbers is a take-off on alpha/beta...  the basic idea
is
to find the nodes that have the most liklihood of affecting the scores
at
the root...  so that you try to avoid going down a path where you
occasionally
have only one really good move.  Because if that one good move turns out
to
be bad, you are stuck...  It's an interesting approach, with a couple of
good articles in the ICCA journal, one by Jonathan Schaeffer in fact.

"conspiracy" comes from the way a group of nodes deeper in the tree
"conspire" to affect the value of a node nearer the root...



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.