Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Exact value of second best move in AlphaBeta with aspiration Window

Author: Omid David

Date: 09:13:06 08/01/02

Go up one level in this thread


On August 01, 2002 at 04:50:23, Martin Bauer wrote:

>Hello all,
>
>In want to detect position at the root node, where the best move is _much_
>better then all other moves. For Example, only 1 possibility for recapture the
>quenn.
>
>I am doing an AlphaBeta Search with aspiration window and calculate the while
>iterative deepening the first move with an 2 Pawnvalue Window, all the following
>moves with an zero window.
>
>My problem now ist, that I have an good value for the best move, but all other
>values are almost as good, as the best move, even if I expect the exact value of
>the best move to be _much_ better than all others.
>
>I think that comes from AlphaBeta, where it is enough to be lower than
>bestvalue, no interest in "how much", am I right?
>
>I there a poosibility to get a better Value at the root node for secondbest move
>with a specific searchdepth?
>
>while iterative deepening at the root node (for distance  := 1 to
>SearchDepth...) I want to ask something like this:
>
>IF (distance = 7) AND (BestValue < SecondBestValue+3*PawnValue) THEN dont waste
>to much time in thinking...
>
>But as I described, even if there is only 1 possibility to recapture a queen the
>values are close together, for exaple 900, 897, 995 ...
>
>any Ideas?
>
>Bye
>
>Martin

In order to get exact values for every root move, you have to use pure MiniMax
there (no AlphaBeta). Of course this will require more search effort, but the
information it yields can be used to compensate that. In addition it will enable
a better move ordering at root.

Omid.



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.