Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: WAC test suite statistics for Mchess 7

Author: Robert Hyatt

Date: 04:54:12 01/15/98

Go up one level in this thread


On January 15, 1998 at 01:06:33, Jouni Uski wrote:

>I run whole test suite (manually) for Mchess 7 in 486 PC, but results
>are scaled to MMX233 level! Here it is:
>
><5 s:  solved  284
>
><10 s: solved  290 (41,80,92,287,296,297 additionally)
>
><60 s: solved  298 (2,71,100,141,152,163,196,265 additionally)
>
>230 is totally impossible, in 235 Mchess wants to play Qe4 with
>+6,xx score after couple of hour - so may be it is side solution ?!

this is possible, as +6 is the score I get for the solution move.  Thru
depth=6, Qe4 is best with a score of +2.5, but at depth=7, after 1.84
seconds, it goes to +3.2 and stays.  At depth=9, 19 seconds Rxf7 fails
high and after one additional second produces a score of +6.  This
doesn't
change over this nor the next iteration.  So if Mchess gets +6 for Qe4,
that seems just as good.

I'm going to make a new suite from the ECM test suite Bruce mentioned.
800+
is too big, so I'm going to cull those that are solved in 1 second or
so.  I
also want to cull a few that Crafty misses, because the move it likes is
+12,
and I'd say +12 is a winning move, even if there is a move that might
lead to
+15.  So I want to cull positions with ambiguous solutions, and end up
with
a set of 300 or so that are more difficult, but not impossible.  I ran
this
on Crafty at 20 secs/move and got 569 correct.  At least 10 it missed I
don't
agree with, because of huge tactical scores for what is supposedly the
wrong
move...



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.