Í want to make some people happy with my engine!
An interview with Richard Pijl, creator of The Baron.
by Eric van Reem
Why on earth does a programmer want to write an alternative version
of his chess program? Programming a chess engine is difficult enough,
so why spending even more time on a Chess960 version? The answer is
easy, according to Richard Pijl, programmer of the strongest Chess960
chess engine in the world: “More and more people asked me if I
could write a Chess960 version of The Baron, and I thought: why not?
Let´s make some people happy!”
Computers in Chess Classic tournaments
In several Chess Classic tournaments computers have played an important
and innovative role. The Dutch program Fritz from Frans Morsch has played
several exciting matches against top players like Shirov and Anand in
Frankfurt, GM Artur Yusupov played two shuffle chess matches against
Fritz in 2000, and one year later Pocket Fritz, the strong chess program
that is running on a Pocket PC Pocket played its first games in Mainz
against Peter Leko and Michael Adams. One year later we have witnessed
two experimental Advanced Computer Chess Matches with Vishy Anand and
Peter Svidler. This year, the Chess Tigers had the brilliant idea to
organise the first ever man against machine Chess960 match. GM Levon
Aronian and the Dutch program The Baron, a creation of Richard Pijl,
will play two games on August 4. Since it is the first Chess960 match
between a GM and a strong computer computer chess program, it is impossible
to predict the outcome of the match.
Computer chess and Chess960
The number of postings in the Computer Chess Club (CCC), in which top
programmers, amateurs and computer chess lovers vividly discuss about
computer chess in all its variants, to be found under www.talkchess.com,
increased considerably after the first Chess960 tournaments and matches
in Mainz. Programmers and chess players alike were interested in the
new chess variant and asked if a computer chess program was able to
implement the castling rules in a chess program. However, there was
no computer program that was able to play the castling rules in Chess960
correctly, but about two years ago the first Chess960 chess programs
Betsy by Landon W. Rabern from the USA and Chispa by Federico Andres
Corigliano from Argentina became available. At the moment there are
six chess engines that are able to play Chess960: the two above mentioned
engines, Frenzee by Sune Fischer (Denmark), Herrmann by Volker Annuss
(Germany), Gothmog by Tord Romstadt (Norway) and the strongest Chess960
engine at the moment, which will play against Levon Aronian in Mainz
is The Baron, by Richard Pijl from The Netherlands. Everybody can try
these engines and use these fine programs to warm-up for the FiNet Open,
because all these programs can be downloaded for free! Go to www.playwitharena.com
to download the free interface that supports Chess960 and download the
six engines from the same site. The Arena interface and the available
Chess960 engines show that it is possible for chess programmers to implement
the specific castling rules without too much difficulties.
However, professional chess programmers seem to be afraid to implement
a Chess960 variant in their engines and there only a few interfaces
that support Chess960. However, Stefan Pohl from Germany shows a chess
computer tournament table for Chess960 with professional chess engines.
How is that possible, you may ask, since chess engines are not able
to exactly fulfil the Chess960 requirements, as they have problems with
the full castling rules. Pohl found a solution: he uses the 15 Chess960
positions in which the king and rooks are placed on their usual squares
and only the queens, bishops and knights are shuffled. But with this
the engines can castle normally, so for the resulting 15 set-ips it
is real Chess960. The engines have to play at least 600 games and at
the moment Shredder leads the table, followed by Hiarcs. Not surprisingly,
these are two engines that have implemented a lot of chess knowledge,
and the results show that these are the best chess playing engines around
at the moment. An interesting experiment and hopefully food for thought
for professional chess programmers.
The tall chess programmer (2.06m) lives in Teteringen in The Netherlands
together with his wife Cantia and his three daughters, seven year old
twins Tessa and Merel and six year old Renske. Pijl studied Computer
Science in Delft and now works for Ericsson as a network designer for
mobile networks. Pijl plays chess himself and his rating is 1809 at
the moment. He spends about 20 to 30 hours per week programming his
engine. He never played a serious Chess960 game but he wants to play
the FiNet Chess960 Open in Mainz to gain some experience. His chess
program The Baron played several tournaments in The Netherlands and
always scored good results. Finding a name for a chess program is not
such an easy task, but Richard Pijl thougt that it would be nice to
implememt the name of his chess club “De Baronie” in it.
A few months after he baptized his program “The Baron” he
wanted to find out how populair his program was by typing “chess”,
“computer”, and “baron” in Google and to his
surpise he found plenty of sites about Baron von Kempelen, who invented
the first chess computer “The Turk” in the 18th century.
“That was even a better reason to call my program The Baron”,
Pijl smiles. The Dutchman has written informative articles about computer
Chess960 in the Dutch computers chess magazine “Computerschaak”
and the next step for his program is an exciting one: On August 4, “The
Baron”, will play a two game match against Levon Aronian in Chess
960. How Richard Pijl thinks about this his world premiere can be read
in this interview.
E.R.: Why did you decide to write a program that is
able to play Chess960? Where you looking for a strong opponent for yourself?
R.P.: I never played a complete Chess960 game myself,
so that was not the reason for creating a Chess960 program. I think
I was just curious. Just before I made the Baron aware of the Chess960
rules, Arena added support for Chess960. That naturally created some
noise in a number of internet fora, and I was asked to add support to
my chessprogram. After reviewing the rules it didn’t look like
a big change to me so I guess it seemed an easy way to make a few people
The Baron’s opponent, Levon Aronian, won the Chess960 open at
the CCM3, which was the strongest Chess960 tournament ever held, and
won the right to challenge reigning Chess960 world champion Peter Svidler
for a title match. Levon might be the next official World Champion in
Chess960. Do you think the Baron has a chance against one of the strongest
Chess960 players in the world?
R.P. (Picture): It is hard to tell as the Baron did
not play many GM’s before. Levon is not just a GM, he is a very
strong GM. Considering the difficulty computer programs have against
the top humans, I can expect that the Baron will have quite a few difficulties
too, as it is not at the same level as e.g. Fritz, Shredder or Junior.
There are a few things different in this match than in previous human-computer
matches though. The time control does certainly favor the computer.
The faster, the better. And the lack of an opening book will mean that
both Levon and the Baron will have to think and use precious time from
move one on.
E.R.: How did you prepare the Baron for this match
against Levon Aronian?
R.P.: At first I started to play a few Chess960 matches
against other programs. Fortunately for me there was a new, strong program
from Norway (Gothmog) that made the Baron sweat for its points. In fact
the Baron lost the first few matches. I noticed that the use of an opening
book has made me lazy. The Baron had a big difficulty in developing
its pieces. So I improved the opening play of the Baron. After this
I did not play many chess960 matches anymore. Instead, I focussed on
correcting errors, adding endgame knowledge, etc. This resulted in a
nice strength improvement of the Baron.
A few days ago I played another Chess960 match with the Baron against
the latest available version of Gothmog. This time the Baron won convincingly.
Another thing worth mentioning is that a few weeks ago I noticed that
a human player, rated a few hundred points lower than the Baron on ICC,
started winning games regularly. It turned out it used a specific type
of attack that is notoriously difficult for computers to calculate (the
trojan attack). So I made the Baron aware of this type of attack and
since then that human player did not win a single game against the Baron.
E.R.: The Baron will play this match without support
from an opening book and without endgame tablebases. That means that
only the program itself decides what move will be played. What is your
opinion on this?
R.P.: Playing without an opening book is the whole
point of Chess960. The computer will have to start thinking from the
start. This is a big difference with normal games. Playing without endgame
databases does not have a big impact on the game. In most cases the
chess programs are able to assess if a basic ending can be won or not,
and are able to win that ending. As I’m regularly testing the
endgame knowledge without using endgame databases I don’t think
this is a big problem for the program.
E.R.: What is the difference between Chess960 and
normal classical chess for a computerprogram like The Baron?
R.P.: Since I didn´t have to concentrate on an
opening book my effort was concentrated on improving middlegame and
endgame play. But there are more differences. In standard chess the
development is completed when the rooks are mobile. In Chess960 that
is not necessarily the case. I've seen a lot of games where the queen
started at a1 and never left that square because the Baron didn't know
that also the queen needs develoment to e.g. a square such as c2 or
d2. When entering the middle game the differences become less, especially
when both sides castled.
E.R.: What is your prediction for the outcome of the
man vs machine match?
R.P.: If this would have been a match on long time
control I think The Baron would struggle not to lose the match. Perhaps
even struggle to win a point. But in rapid/blitz things are different.
Already years ago, in the match of Anand against Rebel, it was clear
that computers were becoming too strong for even super-GM’s on
blitz time controls. Currently, computers are considerable faster, so
I think the Baron has a realistic chance to draw or even win the match.
As I don’t like the sequence of drawn man vs machine matches,
I’ll predict the outcome to be 1.5-0.5 for the Baron.
E.R.: Do you think that the fact that there is no
opening theory in Chess960 favours the human mind over the machine?
Or will the machine benefit from the fact that the human cannot rely
on moves that were learned by heart.
R.P.: There is no opening theory. Even if there was,
the Baron will probably play different moves. Both will have to think
from move 1. And the game can already be exiting at move 2. So I think
the machine benefits from this the most.
E.R.: What are your plans for the Baron?How do you
plan to further improve the engine?
R.P.: I started work on the Baron three years ago
and I basically had to learn how to create a chess program while doing
it. I made a lot of mistakes during programming and because of that
I learned a lot. But many of the mistakes I made are still part of the
most basic structures of the Baron. The way my evaluation function,
where the chess knowledge resides, works is far from efficient. The
same is true for other parts of the Baron.So I decided to redo major
parts of the Baron during the coming winter. I hope that revised version
will be ready to play during the International Computer Chess tournament
in Paderborn (IPCCC) in February 2005.
E.R.: Will the “Aronian-engine” become
available after the tournament in Mainz for the Chess960 lovers?
R.P.: I still like working on the Baron and improving
it. And I like sharing it with others. That is why a new version of
the Baron will be made available for download after the Chess Classic
Mainz. It is tradition that a new version of the Baron is created after
each major tournament that the Baron plays in, and CCM is no exception
Note: The latest version of The Baron can be downloaded from http://home.wanadoo.nl/richard.pijl/