Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rybka and Crafty tie for 1st...

Author: Robert Hyatt

Date: 15:58:16 02/25/06

Go up one level in this thread


On February 25, 2006 at 14:52:05, Dan Ellwein wrote:

>just a guess...
>
>only a mere guess...
>
>:)


Crafty probably won't do exceptionally well this event for several reasons...

1.  It is certainly beyond experimental at the moment.  New search ideas,
evaluation in mid-rewrite, etc.  None of those bode well for a decent result
here.  I only entered to get some data on longer games about search depth,
tactical oversights of new pruning, etc.  So far the search looks pretty good.
But there are holes in the eval where I ripped things out that are glaring.
Today's round 3 vs Fruit was an example, castling kingside and getting into a
big bind was wrong, but the current eval encourages that until it is fixed
later...

2.  No book preparation at all.  Mike made a book from high-level games, but
that leaves more holes than you find in Swiss Cheese, because the book can
become too narrow, and the opponents sometimes exploit popular choices with
cooks.

3.  No special hardware.  Mike's box looks pretty good, but I've seen some 2.5M
nps searches from it which is surprising seeing that my dual xeon is usually
doing 1.5-2.0M...

4.  Had some interesting glitches early on.  I think he started with Arena but
Arena insisted on turning pondering off for unknown reasons.  He finally
reverted to winboard and that went away...

This is just for data, mainly.  We hope to show up at the WCCC with something
significantly better than we had last year.  So far, it looks like that is going
to happen.  But it will take time...




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.