Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: about bug in fruit.2.2(I sent bug report to fabien few minutes ago)

Author: Martin Giepmans

Date: 07:31:26 12/02/05

Go up one level in this thread


On December 02, 2005 at 09:49:14, Uri Blass wrote:

>On December 02, 2005 at 03:16:35, Martin Giepmans wrote:
>
>>Of course only Fabien can know, but a possible explanation is that Fruit uses
>>extensions that depend on game history, for example a recapture extension for
>>recaptures on ply 1.
>>When you enter the game there is a recapture (e5xf6) on ply 1 but when you enter
>>the position there is none. This results in a different search tree.
>>
>>Martin
>
>When you enter the game e5xf6 is not a recapture.


True, according to the strict definition of "recapture". But in my engine I
sometimes (under certain conditions) extend these what could be called
"generalized recaptures" (capture to to-square of preceding move).
I don't know what Fruit does, though.


>A recapture is a capture that is reply to capture and not a capture that is
>reply to pawn move.
>
>I know that there are programs that can extend recaptures but this cannot
>explain different analysis.

>
>I also usually do not expect different order of moves to lead to significant
>different evaluation but only to different number of nodes with the same pv.
>
>history pruning may change the result in one depth but the result continue to >be different also later.

A small difference in the search tree can make a big difference later. Other
hash entries are overwritten, add search instability and you can get a very
different result.

But I am speculating of course. No answer from Fabien yet?


>
>Uri



This page took 0.01 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.