Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Pawn Transposition Table

Author: Robert Hyatt

Date: 11:39:34 01/13/98

Go up one level in this thread


On January 13, 1998 at 12:54:42, Stuart Cracraft wrote:

>On January 12, 1998 at 13:48:02, Robert Hyatt wrote:
>
>>Here's a good sanity check.  I ran win at chess #141 for 45 seconds.
>>I hashed 99% of the pawn scores.  The raw numbers were 4,341,832 probes
>>attempted, 4,327,288 probes successful.  IE Crafty only evaluated some
>>15,000 pawn positions out of 4.3 million calls to evaluate pawns.  This
>>number was similar for every position in WAC (plus others) that I
>>tested.
>>
>>note your hash (pawns) has to be large enough to hold everything,
>>otherwise
>>it is not as effective.
>
>Solution of WAC #141 was slow for me. It took 700+ seconds on a Pentium
>133mhz.
>The pawn tranposition table hashed 98% of the scores.
>
>I know there's been some discussion about #141 before. What kinds of
>improvements help solve this problem? I'm hesitant to go making a lot
>of changes for one problem unless someone says they've helped in other
>areas and are willing to describe them.



This is a hard one, especially if you use null-move, because you toss a
queen early, and that will tend to make every null-move search for the
opponent fail high, hiding the serious mate threats that are there.  The
mate extension Bruce explained helps significantly...

>
>--Stuart



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.