Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Getting a PV using MTD(f)

Author: Andrew Williams

Date: 08:01:00 02/08/00

Go up one level in this thread


On February 08, 2000 at 08:54:22, Tijs van Dam wrote:

>I have been reading about the MTD(f) algorithm that is also used by Andrew
>Williams in PostModernist. I want to try it, but it has some things different
>than a "traditional" search. I am wondering about the following: how do you get
>a PV, or even a move to ponder?
>
>The best move is obvious: it is the last move that failed high at root. But when
>searching that move, all moves in the child position must have failed low. So
>which one of them is the best? Andrew, how do you do this?
>
>My first thought was doing an MTD search with shallow depth from the position
>that follows the root move. Or maybe a PVS search?
>
>Who on CCC has experience using MTD?
>
>Greets,
>Tijs van Dam

I use my hash-table to keep track of the PV. If I've found a move at
a position whose score is > beta, that move ALWAYS goes into the hash-
table. I then rebuild the PV by stepping through these moves. It works,
but often you find the end of the PV being truncated.

BTW in PostModernist I hash moves in the qsearch as well.

Andrew



This page took 0.03 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.