Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hash tables and fails lows.(to Robert Hyatt)

Author: Robert Hyatt

Date: 18:01:34 09/25/98

Go up one level in this thread


On September 25, 1998 at 18:43:17, Inmann Werner wrote:

>You said, using fail lows in hash tables for move ordering is bad. I was not yet
>convinced, but I tried two versions. That convinced me. The version not using
>fail lows for move ordering is about 2% faster.
>
>Werner

there are ways to make this sort of work.  IE suppose at one depth you
store a fail-high (which does have a best move).  at another depth you
store a fail low for the *same* position.  You *could* save the old
fail-high move (since it did go with this position) and use that.  I did
that in a very old version of Crafty, but haven't done this in a long
time...

But if you try to pick a move *at* the point of the fail-low and stuff
in the hash table, I don't see where you can get it from...  As you found...

Bob

BTW, your scientific methodology is *exactly* right.  You should take
*everything* with a grain of salt and try it, to be sure the results fit
*your* program.  Occasionally they won't...



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.