Computer Chess Club Archives


Search

Terms

Messages

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

Author: Fabien Letouzey

Date: 10:55:46 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:

>>Actually in PVS I think of every PV node as a root.  If PV nodes really are
>>different (as they seem to be regarding move ordering), that would justify doing
>>other things differently as well.

>Do you mean that you order moves differently at PV nodes? I guess I should take
>a look at your code. It doesn't sound like a bad idea.

No I don't except for IID now.  But since scout-like algorithms make the
assumption that the first move is the best one (as opposed to "good enough to
produce a cutoff" in null-window nodes), it would make sense in my current
understanding.

>PVS with MTD (f)-like driver. My first engine was MTD (f), and I hope that tests
>will justify using small/zero width windows in at least some situations. Note
>that MTD (f) is a degenerate case of PVS.

I completely disagree here.  I consider PVS to be "move based" (e.g. the scout
assumption) while MTD(f) is value-based (at the root).

Actually I look at PVS as alpha-beta + the scout "trick".  MTD(f) seems very
different from alpha-beta to me (in its way of scanning the tree that is).

Fabien.




This page took 0.02 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.