Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Has anyone tried this in their hash table?

Author: Tony Werten

Date: 00:26:03 08/17/01

Go up one level in this thread


On August 17, 2001 at 01:57:33, Artem Pyatakov wrote:

>I read somewhere a while ago about a technique that was used for a different
>tree-searching game (I am not sure which one it was) and was wondering if anyone
>here has tried to apply it to chess.
>
>Here is the idea that was presented:
>
>As far as I know, most people use two hash tables in their programs - one is
>"always replace" the second one is "depth-is-greater-or-equal replace".
>
>Instead of using the "depth" as a criterion for replacing things in that second
>table, why not use the number of nodes searched as the criterion (i.e. replace
>if the number of nodes is greater or this is an OLD hash entry).
>
>This idea makes sense to me because what we want the hash to do is save us from
>searching the MOST nodes. I think if one combines this with an aging mechanism
>this idea might fly.
>
>Has anyone tried this? What were the results?

It has been tried. It saves a couple of nodes. There are some problems.

Depth only has to take a couple of bits whereas nodes will take at least an
integer. This means your hashtable will be smaller which looses the gain.

Second, if a good move results in a hastable cutoff, the number of nodes will be
low, but the next time you search and this hastableentry is gone, the number of
nodes will be high.

cheers,

Tony

>
>Thank you.
>
>Artem



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.