Computer Chess Club Archives




Subject: Re: calculating the PV using MTD

Author: Dieter Buerssner

Date: 13:31:20 09/11/02

Go up one level in this thread

On September 11, 2002 at 04:08:52, Fabien Letouzey wrote:

[snipped much of the very interesting post]

>Disclaimer (silly attempt to avoid yet another time-wasting flame thread).
>- I do not claim that this method is fast; it is even slower than the
>  traditional way of updating the PV when (value > alpha && value < beta).
>Since the
>semi-PV is updated much more often than a traditional PV (most of the work
>beeing useless, but you cannot know beforehand), its computation is slower.

In my engine, I allways update, what you call "semi-PVs" (for other reasons, I
do not use MTD(f)). When I last checked the performance penalty for this, it was
essentially too small, to be measurable. In cases, where the tail of the PV,
that has to be updated, is long, the engine will have done some enomours work,
to get this, compared to the actual update process.


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.