INTERVIEW WITH RYBKA OPENING BOOK AUTHOR
JEROEN NOOMEN
IR: How did your story
with chess begin?
JN:
At the age of 8 I learned the game from my father. I really liked it and when I
became 12 I attended to my first chess club. My rating improved steadily until
slightly above 2200, but that seemed the end for me. At the end of the 1980’s I
got very much interested in computer chess. The first programs were really bad,
but when I played a tournament in the Dutch town in Dieren, I was amazed how
strong they had become. Then I met with Ed Schröder, the author of Rebel and
when I made my first book for Ed, my addiction to computerchess was born. From
1990 until 1992 I have worked for Hegener+Glaser, the company that made the
Mephisto computers. In the 1990’s the PC made a strong entry and that meant the
end of the dediciated chess computer. Still, I kept on working for Ed’s PC
programs, later also for Chess Tiger of Christophe Theron. Those books were
handtyped, really crazy when I think about that!
IR: How long have you
been making opening books? Can it be a profession? What is so cool in openings
anyway?
JN:
My first opening book was made for the Mephisto Polgar, that must have been
around 1989. So if I am counting correctly, it has been 17 years. I don’t think
it can be a profession, analysing openings the whole day would make me crazy!
But somehow it is cool to analyse modern theory, to find novelties, to improve
on GM games. Probably it has to do with ego, I like it when I can throw a
novelty in a computer game and then show it in New In Chess J. Today the
importance of the openings in GM games are huge, especially with strong
programs as Rybka available. It is simply impossible not to prepare without a
computer nowadays. A wonderful example is the crushing win by Anand against
Karjakin in the Corus 2006 tournament. Great preparation and I think 10, 20
years earlier this wouldn’t be possible. So Capablanca was already suggesting
the ‘draw death’ of chess, but when you look what an enormous jump in analysing
power the last years has given to the chess world, I think it is only the
beginning.
IR: What is fun for you
when you make opening preparation for Rybka engine?
JN:
Absolutely. It was my first available book that was not handtyped, so I had to
start from scratch with a sample of important games. Then the real work begins:
adding priorities, testing, adjusting the book. That cycle repeats itself
several times. You keep on improving until the scores are excellent. Of course
it is great to have a strong engine like Rybka available, as it has no trouble
with any line I chose in the book. It is a kind of luxury problem: Rybka plays
almost any opening well J.
IR: How much time does
one takes to make a decent opening book?
JN:
It depends. When I look at the Rybka book, it took around 1 day to select the
games, then 10 minutes to generate the book. Adding priorities is the most
difficult and tough job, I took around 1,5 months to complete that. The testing
and adjusting took another 6 to 8 weeks, only then I was satisfied with the
results. A good book can probably be made in a shorter period, but it depends
how you define ‘decent’. Of course you can stop after generating the book by
using 1 million games, then play a thousand blitz games and use the learning to
adjust the priorities. Probably you might have something decent already, but I
wouldn’t give my stamp of approval on that book!
IR: Do you prepare some
special novelties for Computer Word Chess Championship in Turin?
JN:
Yes I did, but unfortunately none of them appeared on the board. It is the same
as with analysis: in some games the most beautiful lines are found in the post
mortem, but they never appear in the game itself J.
IR: Do you have a
special very private book for tournaments?
JN:
Yes I have. It is completely different from any known book and its development
continues all the time. I work on it for 2,5 years now, but it will take
another 2,5 years before I will be really satisfied with it.
IR: Do you change book
continuations during the tournament depending on various opponents and present
tournament situation?
JN:
Yes, that is very important. In general I prepare several lines before a
tournament, but all is depending on the official tournament situation. If you
intended opening A for a certain opponent, a ‘must win’ situation can change
your mind. A good example is the game Junior-Rybka from the WCCC 2006 in Turin,
where Rybka had to play for a win. In that game I changed the intended line I
prepared before the tournament.
IR: What is more
reliable: Rybka positional evaluation or GM’s commentary?
JN:
Interesting question! Of course it depends on the lines of the GM are computer
checked. Normally I only trust a line when I have doublechecked if it is right.
There are many mistakes in chessbooks, so as I rule I never take a given line
as the truth. I always check them with Rybka. IMO Rybka’s positional evaluation
is excellent, so I use it very often to check lines. The best attitude for GM
analysis is: take a fresh look at them, check the lines with Rybka and then
come to a conclusion!
IR: What kind of
openings Shredder is good on? When Rybka plays Shredder do you pick some
special lines? How it was in Leiden?
JN:
Shredder is good in king’s attacks, in positions where it has the initiative.
Its evaluation gives high scores for that type of positions. I really like what
Stefan Mayer Kahlen has developed in his program concerning these types of
position. Of course I have special preparation for Shredder, as I have with all
strong programs. In Leiden I didn’t want to reveil anything (the WCCC was still
to come), so I chose a quiet opening line. In the WCCC I was really surprised
to see the Urusov gambit, but I wonder if this really was intended by the
opponent. That’s the mystery in openingbook preparation: sometimes an offbeat
line can give more success than a well known line!
IR: How engines tend to
play in openings by themselves?
JN:
If you start a computer game by move 1 and let them play without book, then I
think most players will agree you aren’t watching a 2900+ Elo game! But that’s
quite OK with me, as otherwise I would lose my job as Rybka book author J.
IR: Do you believe each
engine has its own style and because of that will score better in particular
openings?
JN:
Yes, each engine has its particular strengths and weaknesses. It is the job for
the book author to play into lines the program plays well and to avoid stuff it
plays badly. Sometimes this is very difficult, as it is impossible to take
everything into account. With Rybka it appears different, as I haven’t seen a
line yet in which it scores much worse than in the normal lines.
IR: What is Rybka
engine style anyway? Perhaps she is just only a very fast and efficient calculator?
JN:
No, when I see my own testgames, I simply see Rybka outplaying its opponents.
Its evaluation function is superior to the other programs. Of course Rybka is
tactically very strong, but that’s not the only thing. It plays a nice
positional type of chess, with a very secure and accurate evaluation. And when
the situation is ripe, it will take the chance for tactical blows immediately.
IR: What is a
difference between chess human trainer and opening book author?
JN:
When teaching openings to human players, you’ll have to focus on the ideas of
the openings: patterns, plans, good squares for the pieces, things to avoid and
so on. With a book for a chess program you can only ‘talk’ in concrete
variations. I cannot tell Rybka ‘watch out for Rxc3 in the Sicilian’.
IR: Is Rybka better in
French or maybe in Sicilian? Do you play games to check it or do you use your
chess intuition?
JN:
Rybka likes many lines. It plays the Sicilian excellent, but also the French.
Of course intuition is important – also in making openingbooks – but the real
test is playing testgames. I have experienced many times that a top GM says
‘this line is very good for white’ but when the engine scores 2 out of 10 with
it, you have to face the facts and reconsider.
IR: Could you comment
on table bellow?
JN:
Hm, it clearly shows that Rybka performs well with any opening line. And that
it has big, big scores against the competition, because of which it is no
surprise Rybka leads the computer rating lists! Scores above 80% is a kind of
luxury problem for a book author: what should I prepare!? Anything seems to
do......
Rybka 1.1 32-bits scores with white and black color in
10 majors openings
IR: How chess opening theory will develop now when
everybody had access to computers and engines like Rybka?
JN: I think there will be huge improvements in the
existing theory. Maybe in 10 or 20 years some lines will be unplayable or
refuted. On the other hand I see that chess is such a rich game, that even in
offbeat lines there is still a lot to discover. We are only at the beginning of
the giant computer check of the current theory. Maybe in 10 years all the GM’s
will play 1.d4 b6 and then we have a new field to explore and to discover.
Chess will never be dead, thanks to the computer it will evolve to an even
higher standard.
IR: What do you think about Fischer random chess?
JN: Interesting
experiment, but not exactly my cup of tea! One of the organizers of the Mainz
tournament asked me to come to the FRC tournament that is being held, adding
‘don’t worry, you need not take your openingbooks with you’ J. So from move 1 the calculating starts, very
boring for me!
IR: Deep
Sjeng had really nice tournament in Leiden with 7,5 form 9 clear second place,
but it seem that Rybka win with him pretty easy having advantage already from
opening. Could you comment on that game?
Najdorf Defence B80
Leiden 06.05.2006 Round 4
1.e4
c5 2.Nf3 d6 3.d4 c:d4 4.N:d4 Nf6 5.Nc3 e6 6.Be3 a6 7.f3 b5 8.Qd2 Nbd7 9.g4 h6
10.0–0–0 Bb7 11.h4 b4 12.Na4 Qa5 IMO this is a risky line for Black, GM’s avoid
it at present time. 13.b3 Nc5 14.a3 Rc8 15.a:b4 N:b3+ 16.N:b3 Q:a4 17.Kb2
d5 End of book for Sjeng, far too
early for such a sharp line! 18.c3 d:e4 19.Ra1 Qd7 20.Q:d7+ K:d7 21.Rd1+ Kc7
22.Bf4+ Kb6
WHITE
TO MOVE
Black
suffers in this line for some reasons: his king is not safe, a6 is weak, his
king's side is not yet developed and the white pawns c3 and b4 will become
strong. This is enough to let white sac a pawn with 23.fxe4! A novelty,
played from the book. Khalifman-Gelfand, Corus 2002, ended in a quick draw
after 23.Be3+. According to my analysis 23.fxe4! gives white the advantage. All
captures on e4 and g4 give white a clear edge. But what program will refuse the
pawn!?
The game continued: 23.f:e4!
B:e4 24.Rg1 Ra8 25.Be5 h5 26.g5 Ng4 27.Bd4+ Kc7 28.Be2 Rh7? Already the losing mistake. The rook is completely
out of play here.
WHITE TO MOVE
29.Rgf1
First Rybka move. The opening book gave her a nice
advantage 17 ply +0.39 – white has serious initiative for a pawn. It may worth
to notice very passive position of black rook on h7. 29...g6 30.Nc5 B:c5
31.B:c5 Bd5 32.Ra1 Kb7 33.c4 Bc6 34.Ra3 Kc7 35.Rd1 Kb7 36.Rd6 Rhh8 37.b5 a:b5
38.c:b5 Bh1 39.Rd7+ Kc8 40.R:f7 The pawn is won back and Rybka has a winning
position. I really liked the way in which she smoothly outplayed Sjeng in this
game. Somehow a ‘perfect game’: Novelty, better position and giving the
opponent no chance in the remaining part of the game. 40... Ne5 41.Rf1 R:a3 42.K:a3 Bg2 43.Rf2 Nd7 44.Bd6 Be4
45.Rf7 Rd8 46.Bc4 Bd5 47.Kb4 B:c4 48.K:c4
BLACK TO MOVE
Material
on the board is equal, but white pieces dominate black ones. 48…Re8 49.Rg7
e5 50.Kd5 e4 51.b6 Rd8 52.Bc7 Re8 53.Kc6 Nb8+ 54.Kb5 Nd7 55.Bd6 Rd8 56.R:g6 Nb8
57.B:b8 1–0
IR: Dear reader, you may wonder how to win in Sveshnikov with white in
35 moves?! Easily… J
The only thing you have to do is to get a dangerous passed “b” pawn just
after opening!
1.e4
c5 2.Nf3 Nc6 3.d4 c:d4 4.N:d4 Nf6 5.Nc3 e5 6.Ndb5 d6 7.Bg5 a6 8.Na3 b5 9.Nd5
Be7 10.B:f6 B:f6 11.c3 Bg5 12.Nc2 Rb8 13.a4 b:a4 14.Ncb4 N:b4 15.c:b4 0–0
16.R:a4 a5
WHITE
TO PLAY
17.b5!
Black a5 pawn is weak, and it looks like white will
manage to castle and consume it freely. 17…f5 18.h4 Bf6 19.Bc4 Kh8 20.Qe2
Bd7 21.e:f5 Rc8 22.0–0 B:h4 23.Bd3 Bg5 24.Rfa1 Rc5 25.Nc3 B:f5 26.B:f5 R:f5
27.R:a5 Rc8 28.Nd5 Bh4 29.g3 Bg5 30.b6 Rb8 31.Nc7 Qe7 32.Ra8 Rff8
WHITE TO PLAY AND
WIN
33.Qf3! Game is over, there is no way black can prevent b6 pawn from queening.
If 33...R:f3 then 34.R:b8+ Rf8 35.R:f8+ Q:f8 36.Ra8+- .
Black
played 33...h5 with idea Rxf3, but had to give up after simple 34.R:b8
R:b8 35.Ra8 1–0
JN: You’re
welcome, it was a pleasure to answer the questions!