Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Promotion and horizon, evaluation or extensions?

Author: Andrew Williams

Date: 05:15:29 06/28/99

Go up one level in this thread


On June 27, 1999 at 16:05:46, Brian McKinley wrote:

>Hi,
>
>I and the programmer of WildChild on ICC.  This program was designed to play
>loser's chess. I am now encountering some challenges in trying to make it play
>normal chess at a reasonable level.  I know I can make some quick wins by adding
>an opening book and endgame tables, but I am holdig off until it plays
>reasonably well with out them.  I think they could hide problems.
>
>One of the issues I am struggling with is promotion and the horizon.  WildChild
>recently lost a game to JSBach because it didn't see a pawn
>promotion soon enough.  See PGN below.
>
>What I am trying to determine is if I should try to protect against this through
>extensions, or in the evaluation.  If in the evaluation, what other factors
>besides the location of the pawn need to be accounted for?

I try to use both. I extend one ply in the search if a passed pawn lands on
the 7th, a quarter ply if it lands on the 6th and an eighth of a ply if it
moves at all.

In assessing passed pawns, I check to see if it is safe where it is and whether
it can advance safely. If the enemy has no pieces, I also check to see if the
friendly King is in the square of the pawn. What sorts of things do you look
at?

Could you give the game in PGN? My program can't read the raw ICC output.
To get ICC to send the game to you in PGN, you can do "set mailformat 1".


Andrew Williams


[I've chopped out the rest of Brian's message]



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.