Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: questions.

Author: Dan Honeycutt

Date: 11:37:23 03/24/04

Go up one level in this thread


On March 24, 2004 at 11:30:49, Daniel Shawul wrote:

>Hi
>
>Is it too costy to add checks only for captures in quiescent.
>In my engine non capture checks decrease nps significantly.May be it is because
>it is hard [at least for me] to generate only checking moves.But the capture
>checks seem to work good in test suites and practically no decrease in nps in
>actual play.Has any one have the same experience.
>
>Another completely different question.
>Can attack tables efficently replace the conventional SEE routine.
>Ed says in his paper he uses a table look up scheme to find hanging pieces.
>But what about pinned attackers? [which the conventional SEE handles well].
>Also i can't see how a table indexed by the two 8-bit variables give an
>approximation of the SEE outcome.
>for example if rebel's wb[sq] = 110 11000  [3 white attacks by pawns and
>knights]. How do i know whether it is 2 pawns and 1 kinght or 2 kinghts and a
>pawn attacking the square?
>I think using this method for see will make the already inaccurate SEE worse.
>Am i right?
>
>daniel

I use Ed's scheme slightly modified.  i do 2 bits pawns, 2 bits knights &
bishop, 2 bits rooks, 1 bit queen, 1 bit king.  downside - i have to fetch total
attackers from a 256 byte table rather simple wb[sq] & 7.  upside i know if its
2 pawns and a knight (01 10 00 0 0) or a pawn and 2 knights (10 01 00 0 0).

Dan H.



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.