Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How to manage search depth in limited time?

Author: Robert Hyatt

Date: 05:53:26 03/16/01

Go up one level in this thread


On March 16, 2001 at 05:12:49, Matt McKnight wrote:

>The best way to get the most searching done within the time limit is to check
>your time during the search.  If time is up, return alpha and don't go to the
>next iteration.
>



That is dangerous.

If you are at ply=2 and return alpha, the ply=1 search will think it just
failed high.  You need some sort of sentinel that says "time ran out, do
_not_ use the partial result that is passed back up the tree as you unwind
the recursive calls.

>It isn't wasting time to start searching into the next iteration (no matter how
>much time is left for this move) because even searching only 100 extra nodes
>could give you a better move than the one you already have.
>
>That's why you check your time during search and return alpha when time is up.
>
>Matt



This page took 0.09 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.