Computer Chess Club Archives




Subject: Re: MTD(f) and hash table size

Author: Dieter Buerssner

Date: 15:45:27 08/16/03

Go up one level in this thread

On August 16, 2003 at 18:34:26, Dan Andersson wrote:

> Depth new >= depth old is a pretty crappy replacement scheme IMO.

Same experience, here.

>You risk
>filling the hash with worthless nodes.

Yes. And it may be important, even in "normal" alpha-beta search. I guess, most
MTD(f) engines will use 2 scores in the TT (one lower bound and one upper bound)
while perhaps most "normal" searches will only use one score. Perhaps the
problem is less severe, when the engine uses 2 bounds. I did not try it yet, but
it is one thing, I want to try soon.

There are also some subtle points, that will only show in game like situations
and when not clearing TTs between moves. Especially then, the TT might be filled
with much useless information with the wrong bound. Some aging scheme should
help, but even then, there are traps. For example, an engine might update the
age (too) fast, for some successful probe with the older age.


This page took 0.04 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.