Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: position

Author: Uri Blass

Date: 03:58:06 11/05/02

Go up one level in this thread


On November 05, 2002 at 06:40:30, Sune Fischer wrote:

>
>>My first iteration is ply 1.
>>Why do you use ply 2 as the first iteration?
>
>The root move is done seperately from the search, so 2 plies is the least I can
>do.

I also do it seperately but I still can do ply 1.

In the first ply my alphabeta function usually calls the qsearch(except cases
when there are extensions)

 I am considering doing another stage to prepare for qsearch, but may be that
>is getting over technical.
>
>>I can also see cases when pawns are trapped in the 7th rank.
>>
>>I agree that pawns in the 7th should evaluated as more than regular pawn but I
>>think that bonus of more than one pawn is more than enough.
>>
>>There are cases when it helps but the average result is negative.
>
>I know, but I had to make a quick fix because it was simply getting killed for
>under evaluating advanced pawns. A more precise evaluation on the advanced pawns
>should be done, things like - does the opponent dominate the promoting
>square(s), can he block the pawn from advancing further, is his king within
>reach etc.. Important to know for the evaluation, but also a very dynamic score
>upwards of the queen value might not be a bad idea.

I agree that a better idea is to evaluate these factors(something that movei
does not do today)

I also did a fix of it because it was too small but fixing it too much is a
mistake.

I usually look for the minimal change to avoid the mistake and not for more than
it and even with a constant bonus of 0.7 pawns that is in the latest movei I saw
cases when it evaluated pawn in the 7th rank as too much.

I care about games and not about test positions and I do not want it to push
pawns to the 7th rank when the pawn may be trapped.

I think that without more knowledge 1.7 pawn is a good compromise(I am not sure
about the best value but more than 2 pawns seems to me too much).

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.