Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: how long to see this winning line ...BUT! Is it?

Author: Uri Blass

Date: 10:50:11 11/28/04

Go up one level in this thread


On November 28, 2004 at 11:56:30, Dieter Buerssner wrote:

>On November 28, 2004 at 09:34:27, Uri Blass wrote:
>
>>Here is analysis by yace
>>It seems that it has some bug because it see mate score and lose it only to find
>>it again and mate score should never get lost.
>
>This can happen, because of the transposition tables. I don't think, it is a
>bug. Without HTs, I have never seen this behaviour, while now and then it
>happens with HTs.
>
>Dieter

A mate score is practically always correct with no bugs and no hash collisions
that I guess are not the problem so there is no reason even to search lines
except lines that may lead to shorter mate(if hash collision is the problem then
deeper search will not help to prove that the mate score is wrong).

Movei stops to search after finding mate but if I implement continue to search
after finding mate then I will tell the engine that the score cannot go down
from mate in n to smaller score and that it is pointless to search lines if they
have no chance to help me to prove a shorter mate.

I think that it is possible to do it by having a different window after finding
mate that simply does not let the score to drop below mate in n when mate in n
is the previous score(if the score is not better than mate in n you do not care
about the exact score that research will give and keep the mate in n score).

Uri



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.