Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: an idea to test positional understanding of chess programs.

Author: blass uri

Date: 22:49:11 05/24/99

Go up one level in this thread



On May 24, 1999 at 22:42:48, Greg Lazarou wrote:

>Well, I see all kinds of implementation difficulties with the take-back
>approach.
>
>How about something much simpler: do tournaments with fixed-depth or even
>better, "fixed-number-of-search nodes per move" controls instead of time. So
>lets say you allow a max of 1,000,000 nodes per move, now the more intelligent
>program in terms of positional evaluation, extensions etc. gets to spend its
>alloted nodes in the best way to come up with the move.

I see problem with this idea because programs do not use the same definition of
nodes.

Some programs check if a move is  legal before they continue to search so they
need more time for a node when other programs do not check if a move is legal
and only evaluate lines when the king is captured as a loss or draw(in cases
that they discover that it was stalemate).

These fast searchers cannot see stalemate before they analyze lines when the
king was captured.

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.