Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hash Collision

Author: Andreas Herrmann

Date: 10:27:26 12/07/02

Go up one level in this thread


On December 07, 2002 at 06:55:39, David Rasmussen wrote:

>In my incremental move generator, I first check if there is a hash move, and if
>there is, then I check it for pseudo-legality. If it's not pseudo-legal, I
>conclude that there is a hash collision. The position that this move was stored
>from cannot be the same as the current, and still they have the same hash
>signature. In my program when this happens, I exit. I do this because this
>basically never happens. Until now. It was playing a game on ICC, and it
>suddenly exited. I could see from the log that this is what happened. So I was
>wondering: How often does this kind of collision happen for your engines?
>I think Bob Hyatt has mentioned that this happens in 1 of 100 games. It doesn't
>for me.
>
>/David

Hi David,

as far as i can remember in the papers of Dennis Breuker
(http://www.breuker.demon.nl/thesis/index.html) you can find informations /
statistics how often a collision can happen.
In Holmes i uses a 63 bit signature and i store a hash age value. If the entry
is to old i don't use it.
I also check if the moves from the hash are legal before play them, but i havn't
seen a collision now in many thousands of games. Holmes would beep in this case
3 times :) and write it to my log file that there was a hash collision.

Andreas







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.