Endgame Study Solving Competition

[harold van der heijden]

ARVES, the Endgame Circle of The Netherlands and Flanders (named after
Alexander Rueb), recently organized an endgame study solving
competition.
17 competitors had to solve 6 endgame studies in 90 minutes. Points were
awarded for 'key-move', 'full solution' and accidental busts.

Two computer programs competed; DIEP (operated by Vincent Diepeveen on a
200 MHz Pentium) and Chess Genius 3.0 (on an 'old' 386SX/25).

1. Rob Bertholee (NL) 21
2. Marcel Van Herck (BEL) 16
3. Harold van der Heijden (NL) 14
4. DIEP 10
5. Koen Versmissen 8
6. Uitenbroek/Wissmann 7
8. team/Benak 6
10. Chess Genius 5

[chessman]

great..post the postions..thanks

[harold van der heijden]

Something went wrong; my message was at least twice as long..

I will post it again.

[Bernhard Sadlowski]

In article <32B1DD...@pi.net>,

harold van der heijden <har...@pi.net> wrote:
>ARVES, the Endgame Circle of The Netherlands and Flanders (named after
>Alexander Rueb), recently organized an endgame study solving
>competition.
>17 competitors had to solve 6 endgame studies in 90 minutes. Points were
>awarded for 'key-move', 'full solution' and accidental busts.

Could you publish the studies ?

>
>Two computer programs competed; DIEP (operated by Vincent Diepeveen on a
>200 MHz Pentium) and Chess Genius 3.0 (on an 'old' 386SX/25).
>
>1. Rob Bertholee (NL) 21
>2. Marcel Van Herck (BEL) 16
>3. Harold van der Heijden (NL) 14
>4. DIEP 10
>5. Koen Versmissen 8
>6. Uitenbroek/Wissmann 7
>8. team/Benak 6
>10. Chess Genius 5
>

Why was Chess Genius (3 or 5 ?) running on such a slow machine ? A 386DX/25
is about 1/70 of speed of a P5/200. And a 386SX/25 is probably much weaker.
So the computer results are not comparable at all.

If no second P5/200 was available, then one could test Chess Genius before
the competition starts on the faster machine and keep the results secret
until the main tournament finishes. Wasn't this possible ?

Thanks,
Bernhard

(mailed & posted)

--
Bernhard Sadlowski
<sadl...@mathematik.uni-bielefeld.de>

[Vincent Diepeveen]

This is not important. Much more important was the way the computeroperators
wrote things down. Diep solved 5 keymoves immediately, within a second.

2 out of the 6 studies are instantly solved, but i am awarded only few points
for these 2 positions, as an absolute win (score > 25.00 ) seen by diep
is an absolute win. I wrote down: diep finds absolute win, and i wrote
down PV, where writing down a stupid move would have given diep (and
genius) more points. Problem is: a computer wants to prevent mate.
Instead of the opponent allowing to go mate it prefers to do an objectively
better move, where the loosing move (or easy draw move),
is the move you get points for.

Next year, if i get invited with my program, i will definitely beat
the best human.

I will not only write down the best move my programs finds, but i will also
move the pieces of the opponent, doing moves which look promising for the
loosing side (or the try to win side). Finding the best move is something
i hand over to my program. This first time i mainly tried to enter moves
my program found.

Things that went wrong this time:
a) i didn't know that the first 2 studies were so easy that a program finds
it within a second, giving diep more time for the other studies.

b) It appears that thinking for over 10 minutes, current version 12 minutes,
diep would also solved the rookending, where i was so stupid to
analyse 45 on the first 3 endings.

after diep got a fail high (found a move better than the previous)
in that rookending from d6 to Rg2 i thought he solved
it. I should have given diep another ply extra, after which it would
have found Rh6!

I did not do this, i plead guilty.

So seen from an objective point, both Genius and Diep suffered from the
way points are awarded. Diep will NEVER try to go mate earlier if it
sees that it can escape, although a terribly lost ending, it is not
a mate yet.

The real problem was that the operators joined this year for the
first time. Not the programs were the problem, the annotating programmers
were the problem.

>Thanks,
>Bernhard
>
>(mailed & posted)
>
>--
>Bernhard Sadlowski
><sadl...@mathematik.uni-bielefeld.de>

Vincent Diepeveen
--
+----------------------------------------------------+
| Vincent Diepeveen email: vdie...@cs.ruu.nl |
| http://www.students.cs.ruu.nl/~vdiepeve/ |
+----------------------------------------------------+