Computer Chess Club Archives




Subject: Re: Hash Collisions (Dan)

Author: Dan Newman

Date: 12:23:14 02/23/99

Go up one level in this thread

On February 23, 1999 at 14:43:29, Larry Griffiths wrote:

>On February 23, 1999 at 14:33:03, Larry Griffiths wrote:
>>I read your post on hash collisions.  I have tried some hamming distance
>>code and I see little difference when compared to generating random
>>numbers.  I have been thinking of writing code that does what you said
>>where 64 bits must have 32 bits (or 50%) turned on in each number.  I also see
>>where there are many references to the 12 pieces times 64 squares = 768
>>hash codes.   Pawns only use 48 squares for each side, so 32 squares are unused by pawns.
>Forget my reference to the Bishops.
>>768-48 leaves 720 hashcodes for the piece square table.
>>Food for thought?  Errors in my thinking?

That's interesting.  You can generate a smaller set of random numbers and
fill out only those portions of the table that are actually used.  And a
smaller set can be better optimized in the same period of time. (Currently,
IIRC, I use 1024 numbers for the piece-square part since I use Bob Hyatt's
numbering scheme for pieces: P=1,N=2,K=3,B=5,R=6,Q=7 -- which is quite
nice for distiguishing sliders from non-sliders.)



This page took 0.01 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.