Author: Dirk Frickenschmidt
Date: 09:14:34 10/10/97
In Summer '84 - I feel like talking of "summer of '42" :-) - , when I
was student in Heidelberg and still writing articles for CSS (a german
computer chess magazine existing still today), I published a - as far as
I know - then new idea within computer chess for CSS 4/5 (October 1984):
"Das Thema-Turnier", I'm not shure how to translate it, but I will
explain it.
I had five board chess computers play a tournament with the same opening
moves.
These moves were from the Kortschnoi-Kasparov candidates tournament, a
Tarrasch defence successfully used by the later world champion:
1.d4 d5 2.c4 e6 3.Nf3 c5 4.cxd5 exd5 5.g3 Nc6 6. Bg2 Nf6 7.0-0 Be7 8.Be3
c4 9.Ne5 0-0 10.b3 cxb3 11.Qxb3
Now Kasparov continued with the somewhat risky 11...Qb6
Each program played against each with white and black. Results then:
1. Elite A/S 4.6 Mhz 5.5 points
2./3. Constellation 3.6. Mhz 5
Mephisto Exclusive 5
4. Playmatic S 3
5. Superstar 1,5
The position allowed at least two good moves for black:
- The Kasparov move Qb6 was played by none of the programs
- Qa5 (recommended by Pachmann) was played by Mephisto
The other programs tried: ...Bd6 (Elite, Playmatic) - a maybe not
completely convincing, but interesting alternative found by the
computers -, ...Nxe5 (Superstar) and ...Re8,
both rather questionable moves.
All these alternatives were leading to different kinds of games, most of
them quite entertaining.
The idea behind the whole tournament was:
comparing what programs make from a certain (more or well known)
position in more than one move and *not* only considering a key move or
their first move, but their whole kind of play from the start position
on.
There should be *no* forced "only winning" move in this kind of test.
Only the result of the whole game - so the comparable overall perfomance
- is decisive and worth observing and commenting.
After having seen many good and bad testsuites with key moves, now I
return to my old idea (meanwhile also practiced in the so called
"Nunn-Test" for Fritz 5) and will continously create a test set of
thematic positions suitable for chess programs, not only openings this
time, but also late middelgame and endgame positions.
In my eyes this is one of the most interesting areas of computer testing
in the present age of auto-players. And to me the overall performance in
such well known positions is much more informative than long
autoplayer-sessions with very big books and/or dozens of conventional
test-sets (which have their own merits nevertheless!)
*** Now I need your help: ***
Not being able to do all the testing on my own (besides my P133/32Mb
mostly in use I only can produce P90-games on two other machines with
16/24MB and therefore limited hastables), I will continously publish
test positions here, asking for your help.
I hope those of you who are interested will send me their test games
played with these positions.
For each test position I will collect all the games you send to me by
e-mail (or published here) in a database. I will add some more human
sample games/comments played with this position if available. Then I
will send this database to anyone who has contributed to the current
test.
After collecting enough positions for my "play-the-game-test" (not only
from openings, but from late middelgames and endgames as well) and
sorting out others which proved not to be relevant enough I will publish
the whole test as testset with all games played so far in a database.
I hope to supply free access to this database, probably by giving it to
some well known www-pages.
For a first test - and to see how many of you have fun in contributing
games,
I ask you for the sake of some "Heidelberg nostalgia" ;-) to play the
above Tarrasch variation with your programs and on your platforms
(please only use same hardware for both programs and name settings).
Preferred are tests on tournament level 40/120 on either P90 or MMX200
hardware.
(This makes the results comparable to those ot the SSDF-database).
By the way: the same CSS-magazine contains an article about the
micro-championship in Glasgow with photos from people like Kathy
Spracklen and information from the London games of IM David Levy against
Cray Blitz with a nice photo of Bob Hyatt - more hair then :-) - and
Hans Berliner - only as much hair as needed ;-)
Kind regards from Dirk
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.