Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hashing and draw by repetition

Author: Inmann Werner

Date: 01:48:43 08/26/99

Go up one level in this thread


On August 26, 1999 at 04:31:40, Bruce Moreland wrote:

>
>On August 26, 1999 at 03:23:43, Inmann Werner wrote:
>
>>Hello
>>
>>When I began to implement hashing, I tried to use all "tricks" to avoid draws
>>by repetition going into the hash tables, because I thought, this would be very
>>bad.
>>Yesterday I disabled this code and let also the draw by repetition evals into
>>the hash tables, and surprisingly, the program worked better (at the complete
>>LCTII Test)
>>
>>Do you also let "draw by repetition evals" into your hash tables?
>
>It so happens that in mine is it as easy to avoid as to not avoid, so I avoid
>it.  I doubt it matters much.
>
>If you avoid it, you still have path-dependent scores in your hash table, which
>will cause search instabilities.
>
>As far as I can tell it is impossible to avoid this without hashing the entire
>path taken to get to the position, which is insane and likely wrong.
>
>bruce

I did it, setting the eval of draw per repetition to 1, a value my eval can not
produce, and not allowing any 1 to be put in hash or got from.
But through cutoffs and similar things, the "draw per repetition" influences the
hash entrys anyway!?
.
When I let the "draw per repetition" evals be put in hash and used, in most
position there was no change. But in positions, where I stand lost, but there
are chances do come to draw, they got solved much quicker.

???
Werner



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.