The Rybka 4 Book by Jiri Dufek

The Rybka4 opening book is a high-class compilation of opening theory. It was developed for the world‘s strongest chess entity – a 100-core cluster running the program Rybka4, used to maximize the program‘s playing strength. The choice of variations in the openings book is well suited to the style of the program, and the book has been fine-tuned and checked in thousands of games.

The Rybka4 Book includes 17 million positions and over 700,000 games, which were selected from following sources: top human events, correspondence chess games and computer chess games (up to April 2010). And this is only the starting point: the value of the book lies in over 700 original analyses and move choices, which lead to a re-evaluation in many critical points of modern theory. This makes it useful for tournament and correspondence players alike.

The book contains a lot of computer checked analysis never published anywhere before. The author of the Rybka4 book is Jiri Dufek (national master, FIDE 2276, ICCF IM 2579), one of the world‘s leading experts in chess opening theory. Jiri Dufek has been on the Rybka team since summer 2009, and after this date his preparation was used in all official events which Rykba won. During World Championship match between Vishy Anand and Veselin Topalov Jiri worked as second of Veselin Topalov. In his long chess career he won twice Czech Correspondence Team Chess Championship and wrote together with Roman Chytilek book Beating the French Defence (in Czech only).

Interview with Jiri Dufek

Yiri Dufek: When I was six my grandfather taught me how to move the pieces. He was a relatively strong player, I think about 2000 in his old age. However, he never reached the national candidate master title. A year after learning the game I started to play in a local children’s club, and my first trainer was Rudolf Soukal. However he was very busy at the time, and I interrupted my training session for two years. At ten I joined the best local club, Spartak Usti n.L., for a couple of years. At the age of 18 I was rated only 2024, however two years later I made a jump to 2266 and stayed there for a long time. At the same time (in 1992) I started to play correspondence chess, as a member of the "B" team of my city. My first experience was bad, but during the next season I played much better and then got an offer to play in the "A" team. With this team I won the national team championship twice and made one third place.

During this period I met IM Roman Chytilek and we started a long-term cooperation. Together we wrote a two-part book called Bijte francouzskou! (Beating the French), which is now available for download on my webpage, however only in Czech. Our latest success was to win the 4th PAL/CSS Freestyle tournament on Playchess. Lately I stopped play OTB chess and started spend time into analyzing openings and interesting ideas in general. In summer 2009 I got an offer from Vasik Rajlich to prepare a book for the Rybka cluster, and later to prepare the book for the new Rybka. This offer started our cooperation, which led to one peak of winning the MundialChess Freestyle tournament, using his 108 cores cluster during this tournament. At the moment the latest milestone in my chess career was to be the second of Veselin Topalov in his World Championship match against Vishy Anand.

I took my first steps while testing on Playchess a few years ago, but I could hardly call this making a book. I started to prepare a real book for the Rybka cluster during the summer 2009. It is really interesting to analyze unknown positions and ideas from chess books, chess practice or computer games, and to find your own solution. However it is not possible to do this every day for eight hours. Today's chess preparation of players is very deep – it is not surprising that in some games players lose in the middlegame because the opponents simply know the lines for dozens of moves, from the opening to the endgame. This is even worse in computer games: long lines, very often 50 moves long, lead to draws that are similar to prearranged human draws.

Yes, in some ways, but it is hard work too. The first important thing is a good choice of games, because is not possible to make wide books purely by hand. The second thing is to choose the "right" openings – a narrow book can perhaps score pretty well for a short period, but I wanted to add much more information value. This is why I added human openings for the non-tournament book. The third thing is the testing phase, which works in cycles – testing, finding weak points, analyzing and improving bad lines, updating the book, new testing... During this period I used Aquarium 2004 and Chessbase 10 for making the book.

How much time does one take to make a decent opening book?

A lot of time. Finding the right games and generating the book is the work of a few hours. Toying with priorities is a very time consuming task – it took a number of weeks. Testing games is an independent process, but only checking of results, finding lines which need corrections, that is the work for a few days. Last but not least is the phase of finding new ideas and understanding why the engine plays badly in very good positions, from my human point of view. This phase is not only very time consuming, but needs a lot of invention and ideas.

Yes, but for Rybka 4 I moved my priorities away from the tournament book which I used with the Rybka cluster. Generally I can say that the Rybka4 Book includes maybe 95% from the book which I used during tournaments, plus a lot of new material – analysis, up-to-date games, etc.

I want to say for everyone, however I am not sure that this is really true. There are three groups. First the chess engine fans. They will find up-to-date information about their openings. The second group would be everyone who wants something new, new ideas or new setups. And last but not least the book is for chess players, as a highly topical chess guide.

No, I didn't. From my experience, GMs are not so interested in this computer engine stuff, but if there is any feedback I will be happy, and it could be useful for future work.

During my visit in Linares I met Veselin's seconds – Jan Smeets and Erwin l'Ami. After my game with Veselin we spoke generally about chess, openings engines and differences between human chess and the chess engine world. When I returned home I got an offer from Silvio Danailov to join the team of Veselin for his match against Vishy Anand.

I think this question is more complicated than it looks at first sight. Firstly Veselin (and GMs in general) have their own ideas on how to play openings. It is totally different from engines matches. Nobody is interested in long lines which can draw at end, because memorizing these lines is very difficult and time-consuming, without any big chance for success. Secondly I very often checked existing analysis and was looking for improvements or analysed lines which were assessed as important for a match. We ended every time with the same conclusion in our analysis. Thirdly I provided quite a successful IT support for the team.

Yes, of course. I have used ChessBase products for years. I use ChessBase 10 as the main program for storing my analysis and for my work with databases. I found very useful online access to database. The Fritz GUI I use for automatic “Deep position analysis” – for analyzing positions and for testing purposes.

Generally the book is relatively wide. I wanted to prepare two ways to play every opening, but sometimes I used only one, because I thought that was best. But the main repertoire is based with white on the Najdorf with 6.Be2/h3/Be3/Bg5, Caro Kann with 3.e5, Ruy Lopez, French with 3.Nc3 and 3.e5, the Catalan, the classical KID with 9.Ne1 and 10.Be3, the Exchange Grunfeld, Nimzo Indian with 4.f4, etc. For Black there is the Sicilian Kan (e6+a6), which the cluster proved was very good in tournaments, the Sicilian Najdorf and Rauzer with Bd7, the Caro Kann and Ruy Lopez (Berlin Wall and Jaenisch). After 1.d4 there is the Grunfeld and a lot of Slav (Chebanenko 4...a6) and Semi-Slav stuff, same as the Nimzo Indian, where I used my openings from the cluster book.

It's very hard to answer. Probably the most time consuming was the Najdorf, but I am still not 100% sure about the truth in this opening. The most fun I had was analysing the Jaenisch in the Ruy Lopez, because very often not every best move of the engine is really the best. But that is not all. I added about 700 analyses of different openings into the book. They are not often set for the tournament book, but mainly for wider use by the players. In other words there are continuations of sidelines which have improved existing theory, or recent games – look and you will see for yourself.

Sure…

The second part of this interview with novelties and
new ideas from the Rybka4 book will appear in the next fews days.

Rybka in the ChessBase Shop

Order Rybka 4 now in the ChessBase Shop