Computer Chess Club Archives


Search

Terms

Messages

Subject: About testing chess engines

Author: Aloisio Ponti Lopes

Date: 21:34:51 04/04/01


I remember the days I used to play blinfold chess against SargonIII running on a
286-20 MHz with 4 MB RAM and 20 MB Hard-disk, just to improve my "chess vision",
as my school chess trainer once taught me. That was really funny, and it REALLY
worked, I improved my game _a _lot.

Now we have those wonderful machines, Athlon or Thunderbirds and Pentiums
running at > 1 GHz; 256 MB of RAM is afordable, and we can also have 15 GB or
larger HDs, fast video cards, large screen monitors, etc.. I bought the WM-Paket
 and installed the Nalimov tablebases... that's more than 2 GB of hard disk
space that's only for playing 3-4-5 piece chess endgames - something that at the
time I was using SargonIII I could not even dream of.

Now I can test many engines against each other to decide which one is going to
be my next "sparring", or which one I will use to analyse my games, as I can't
pay for a GM to do it for me.

Ok, I'm not a master, but I like Chess. I mean all "variations" of chess.
Bullet, blitz, standard, blindfold, progressive, bughouse, random or whatever
exists, I like it.

No, I'm not a chess programmer, and I'm not a computer expert, but I try to test
 those programs for my own use, and for some of my chess-students (most of them
are children).

So that's maybe a different vision than the "experts" here have.

I don't think it is nonsense testing a chessprogram using a Beowulf cluster of
Pentiums running Linux. There are many other applications to this kind of test,
in science (well, maybe chess is just an interesting algorythm to test those
super-machines). I don't think it is nonsense testing a program using common
affordable machines. I think TESTING is the only thing that I can do to decide
which program I can "trust" more.

So here's my humble opinion about this topic: I like to test the engines
beginning with 1 minute bullet games because I think I'm testing the SOFTWARE,
not the HARDWARE. I don't own a 200 Pentium-Beowulf-cluster running at home! ;-)
I don't even own a dual!(I spent 10 months paying for a license of Windows 2000
Professional, just to get rid of those disgusting 95 & 98SE GPF errors). A
faster software seems to be better than a slower one; also if it has more
knowledge, it will be better than softwares with less knowledge. All that we
want is the best of the two worlds, isn't it? Faster and smarter (fast searcher
with as much knowledge as possible).

Comparing this with humans: I have a friend who is a Fide master. If I set up a
position and show him, he can "see" much more in 1 minute than I can "see".
Maybe I could stand up in front of the board the whole night to see what he
sees. Maybe I couldn't, and ask some friends to analyse with me. Let's say I can
 call 15 friends to analyse the position; maybe at the end of the night, our
analysis won't be as good as the FM's one. I once read that E. Lasker did
something like that in a chess club somewhere. Many chess players could not see
what he could easily see. Capablanca once did it too when he was playing in
Moscow; many "strong" chess players all together couldn't see an endgame path to
win, but Capablanca saw it in a minute...

I know of course that testing engines with more time is important too; but that
makes me remember how foolish my analysis are comparing to my friend's
analysis... well, if the position is just easy, he sees the solution and stops
thinking. If it is complicated, he can write many pages about it, giving
variations and verbose analysis that many often are quite difficult to
"understand": his chess knowledge is much better than mine!

-------------------------------------------------------------------------------

A. Ponti



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.