Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Iterative Deepening

Author: Robert Hyatt

Date: 11:04:49 09/22/98

Go up one level in this thread


On September 22, 1998 at 09:03:45, Pat King wrote:

>
>On September 21, 1998 at 17:58:23, Tom Kerrigan wrote:
>
>>Iterative deepening might be a wash at 4 ply, but I would not like to start a 10
>>ply search with an empty PV array/hash table/history array.
>>
>>-Tom
>
>Yes, but if you're not doing ID, you wouldn't clear out the data from previous
>searches,
>but use it to give you pretty good ordering.  Certainly a deep search with
>-ZERO-
>move ordering would be painful, but  you've still n-1 or 2 deep data from the
>last move.
>It seems ID throws that all away.
>
>Pat

Rather than us trying to explain why/how it is good, or you trying to
understand second-hand, get the source to any program and try it.  IE
in Crafty, you can do a search to depth N using iterative deepening, then
modify iterate.c to stop the iterated loop and just start at depth=N and see
what happens (note that Crafty uses internal iterative deepening as well so it
will not do as badly as a program without this, starting at depth=N).

I didn't believe this would work when I read about it in 1976.  I tried it in
1978 and have not used anything else since... :)



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.