Computer Chess Club Archives




Subject: Re: An MTD(f) question about NULL MOVE searching

Author: Tord Romstad

Date: 10:49:41 07/16/04

Go up one level in this thread

On July 16, 2004 at 13:40:52, Vasik Rajlich wrote:

>On July 16, 2004 at 12:13:25, Fabien Letouzey wrote:
>>As for the "true bounds", Tord mentionned that he used some form of "aspiration"
>>together with MTD(f).  The guess part of aspiration can obviously be applied to
>>any algorithm and he makes some decisions during the search depending on those
>>values.  So I am not sure about the "decisive advantage" although I don't like
>>the idea of MTD(f) at all.
>Maybe Tord found something interesting. In MTD (f) you do have root aspiration
>values, but these don't seem that useful.

The idea is not mine; I learnt it from Andrew.  At the beginning of each
iteration, I assume that the score will end up within a window of +- 2 pawns
centered around the result of the previous iteration.  I use this window
as a basis for pruning decisions.  When it turns out that the score is
likely to end up outside the window, I clear the hash table and restart
the iteration with most speculative pruning tricks disabled.


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.