Interview with world champ chess programmer (long)
Kees de Graaf
Nederland" contains an extensive interview with Ed Schroeder by IM
and journalist Max Pam. Schroeder's program "Chessmachine" became
world champion by winning the tournament in Madrid, last December.
Some of you folks may find it interesting if I transcribe here some
statements made (copyright Max Pam, I guess).
Background: Chessmachine became world champion by beating all of
its contestants, including Hans Berliner's Hi-Tech. In its
commercial version it sells for ~$800 in microcomputer software
shops. Schroeder works on his creation in a shack beside a church
in Deventer, The Netherlands.
Q: What's your driving force?
Ed: Hobbyism. I'm a low grade chess player (Elo 1800) and a
mediocre programmer, but I wanted to do something special. I had
some ideas, which I implemented. The first time I lost from my
program I experienced it as a triumph.
Q: How strong is Chessmachine now?
Ed: About Elo 2350. However, in Madrid it scored a tournament
rating of 2450. That may have been a lucky shot, but I've been
wondering how the program would perform running on Deep Thought's
hardware. A doubling of the speed of the PC Chessmachine runs on
gained me 60 Elo points. Deep Thought is a hundred thousand times
as fast as that. May be Chessmachine running on the Deep Thought
hardware could eventually score in the GM ranges.
Q: Has your program a certain playing style?
Ed: It's rather aggressive, but in a responsive way. My ideal is
Tal's style, with its intuitive appearance. Chessmachine is
particularly good at positional sacrafices. In Madrid it sacced
a pawn versus Hi-Tech, and I thought oops a bug, this is
completely idiot. When we analyzed the game it turned out to be
a very interesting and correct conception.
Q: What's special about the program?
It has an extremely efficient handling of recurrencies in tree
structures, which is based on a mathematical concept I found a
long time ago. Furthermore, the evaluation function focusses on
pawns and pawn structures. In the Cologne 1986 tournament it
played away the world championship in the last round because it
mishandled a free pawn. I decided that that would never happen
again, and sought and found the definition of the actual danger
of an advanced free pawn. It was difficult though, to translate
that definition into computer programming terms, but at the end
I succeeded.
Q: So, what is that definition?
Ed: I cannot give that away, I'm afraid. The competitors are
listening. The only thing I can say is that it is surprisingly
simple and clear. It is a rule, in fact, that probably most
strong players know unconsciously.
Q: You're the world champion now, yet Deep Thought is generally
regarded to be the strongest.
Ed: Deep Thought combines fast hardware with little chess
knowledge. It computes 7 moves deep and additionally another 3
to 4 selectively. I would probably loose a match by 7-3. But if
I could employ their fast hardware, the result might be
reversed. My program has chess knowledge implemented as simple
rules, such as: what is an open file worth, and what is the
meaning of a doubled pawn. Just the sort of basic things that
you learn in chess class at secundary school.
Q: Why didn't Deep Thought compete in Madrid?
Ed: Deep Thought has been bought out by IBM with a lot of fuss.
Since then, no results. IBM has injected millions of dollars
in the project and cannot afford a defeat at this point in
time. Even a second place of this 20 million dollar machine
after a PC with a $800 program would be disastrous. It has
happened before: me playing against a gigantic machine and
winning, and then afterwards have a look at their faces when
I draw the floppy out of the PC!
Q: You're working in a church building, are you a religious
person?
Ed: Yes. I find it surprising that I could achieve all this,
being a mediocre man, and I'm grateful to the lord.
Q: Isn't the game of chess damaged by the existence of chess
computers?
Ed: Nonsense. I think it's right that computers are disallowed
at important tournaments. You don't allow horses at athletics
races, either. You should use the computer to help in finding
the truth in certain positions, that is what it's very useful
for.