Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: questions about WAC

Author: Uri Blass

Date: 04:11:57 12/27/01

Go up one level in this thread


On December 27, 2001 at 06:58:41, Gian-Carlo Pascutto wrote:

>On December 27, 2001 at 03:51:22, Uri Blass wrote:
>
>>The only problem that it could not solve in a reasonable time
>>is WAC 2
>>I guess that it needs depth 13 in order to solve it
>>and the estimated time that
>>it needs to get this depth is at least some hours.
>
>Nothing to worry about. WAC 2 is hard, both for search
>and eval. I get it at ply 11-13 depending on the extensions.
>Usually 10-30s on Athlon 1000.
>
>>The hardest problem for it
>>(from the problems that it solved) was WAC 22
>>but I consider it as a positional problem because
>>there are many moves that win a pawn for white to get equality
>>in material.
>
>>Qh5 is one of them and my program found Nxf7 at depth 4
>>but changed it's mind at depth 6 to Qh5 only to change
>>it's mind later again to Nxf7 at depth 7.
>
>I get +1 for white with Nxf7 at 13 ply.
>
>After playing out Qh5 I get g6 and BLACK up 0.70 pawns.
>
>So Qh5 looks a lot worse as Nxf7 for my program.
>
>--
>GCP

I agree that Qh5 is worse than Nxf7 but for positional reasons

If you use only material evaluation both moves lead to equality and

My program has now no evaluation except piece square table
and I prefer to test it to see that it has no bugs
before adding new evaluation terms.

There are other moves that lead to equality from material
point of view and Ba2 that is written as one of the solution
is one of them.

Yace(only material) evaluates Qh5 as 0.00 and later
changes it's mind to believe that it is losing material and
later changes it's mind again to 0.00 so maybe my program
found that qh5 is losing material and I did not check it
(it only gives main line and score in the end of the iteration
and I am going to change it).

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.