🚀 go-pugleaf

In rec.games.chess.computer Tim Benham <timben@dworkin.interact.net.au> wrote:

:>I ran a match at 60 moves in one minute this morning to make sure I
:>hadn't broken something.  Crafty won 7, drew 1.  I will play more
:>later.
:>
:>the very first thing that comes to mind is, was this experimental
:>match set up properly?  IE it was run on one machine.  Was gnuchessx
:>put in "easy" mode to match the ponder=off in crafty?  Did you confirm
:>that only one was running at a time?  IE in linux, while my match was
:>running I used "top" to be sure that I had only one cpu-bound process
:>at a time.

: No, this match wasn't set up properly -- ponder=off and easy mode
: looked too difficult to bother with -- but I don't think that should
: bias the results towards gnuchessx. I did run top during a test game
: and found both programs were receiving about 49% of the CPU whoever's
: move it was, so this isn't the explanation.

: One thing that did make a difference is the broken results keeping by
: xboard 3.6.2. Because of that the scoreline reported in my original
: post was simply wrong. I examined the games and found the correct
: result to be 11-7 in Crafty's favor!

: This still seemed to be more level than I would have expected. I
: guessed that Crafty might be hurt by a too large hash table, causing
: it to swap. So I rematched the two programs with hash=12M (vice
: 16M). After 10 games Crafty scored 1 1 0 = = 1 1 1 1 = : 7.5/10.
: (gnuchessx's lack of chess knowledge made some of the games painful to
: watch -- imagine K+R dancing with k+r for 50 moves!).

hash=12M and hash=16M are identical, as crafty rounds to multiples of
1.5M due to the way it accesses memory for hashing...

and I've seen a few such things happen with drawn endings...



: However, I made a disturbing discovery which may invalidate even this
: result: gnuchessx may lose on time and crafty does not claim! is this
: a known problem in crafty or xboard?

I just fixed this.  I didn't do this in the past because usually
xboard is talking to a chess server, and the server claims the game.
I'm going to check to make sure that xboard won't quit a game when
Crafty calls "time" *if* playing on a server, which could cause
problems...


: -- 
: 		  People who like this sort of thing
:        	      will find this the sort of thing they like.
: Tim J.Benham					timben@interact.net.au

-- 
Robert Hyatt                    Computer and Information Sciences
hyatt@cis.uab.edu               University of Alabama at Birmingham
(205) 934-2213                  115A Campbell Hall, UAB Station 
(205) 934-5473 FAX              Birmingham, AL 35294-1170