Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fruit 2.2's strength

Author: Dann Corbit

Date: 08:03:36 10/26/05

Go up one level in this thread


On October 26, 2005 at 08:24:59, Uri Blass wrote:

>On October 26, 2005 at 06:42:51, Tord Romstad wrote:
>
>>
>>On October 26, 2005 at 04:32:49, Fabien Letouzey wrote:
>>
>>>PV-nodes search results are stored into the transposition table.  What Dann
>>>meant is that transposition cutoffs are not permitted at PV nodes.
>>
>>I still don't understand what Dann meant, but I think it must be something
>>more than this.  Not allowing transposition cutoffs at PV nodes is a rather
>>common thing to do (I've always done it), and it has very little influence
>>on the efficiency of the search.  Dann seemed to be talking about something
>>which no other program except Fruit does, and which should normally have a
>>big impact on the search.
>>
>>Tord
>
>I think that Dann was simply wrong.
>Note that I even did more than it and never allowed tranposition cutoff in
>Movei.
>
>One of the reasons is that my evaluation is dependent on the path and not only
>on the leaf position and another reason is that I do not hash partial depth so
>cut at the same depth may be wrong and cutting only when the remaining depth is
>bigger depth did not change much in the last time that I tried it.
>
>I may try to keep evaluation that is dependent on the path in the new version
>that I may write or not write(I do not promise nothing) but I think to try first
>normal evaluation without partial extensions.

Yes, I was wrong.  I misunderstood an explanation that I read some time ago.



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.