Interview with the programmer of Rybka 4, Vasik Rajlich:
Rybka 4 is much stronger as compared with Rybka 3? What areas have been improved specifically? Is it possible to estimate the Elo of Rybka 4?
Rybka 4 is improved in many areas. The search is faster and more efficient, and the evaluation is more accurate. I have worked on almost every area a little. A significant improvement lies in the understanding of King attacks.
As for Elo - this is a question for the testers. The exact number will depend on the test conditions.
How do you go in the improvement of Rybka?
Normally I take an area and test dozens of small changes, which I held either reject or confirm, based on the results of an automated test procedure that I have developed over the last few years step by step.
The process is incremental - since Rybka 3 there have been over 3,000 unique Rybka versions, and each plays differently.
How much is Rybka improved, when multiple processors are used in parallel?
A good rule of thumb is if you double the number of processors, then Rybka "speed" is increased by a factor of 1.7.
How would you describe Rybka's playing style? Is Rybka particularly strong in certain types of position or openings ? Do you recommend a special opening book for Rybka 4?
Rybka's style evolved over time and I myself can not judge the likely best. What I can say, however, is that Rybka has become more aggressive and tactical with time. Rybka versions up to 2.3.2 were tactical relatively weak compared to other programs , and that's was noticed by many Grandmasters. As for computer chess, GMs are in my experience, usually an interesting mix of ignorance regarding technical questions and a rather clever eye for the weaknesses of each program.
The improvement of tactical alertness was certainly in the last few years an important focus for me.
As for the opening book - author of the Rybka 4 opening book is Jiri Dufek. He has used for analysis and test Lukas' cluster (see chessbase.com / newsdeteil.asp? Newsid = 5444), and this book is much deeper and more accurate than any other which is currently available. It should be noted that the book will be "objective" - Jiri was keen to find the truth and not some variants that are liked by Rybka. This is a deliberate decision of our team. Our goal is to provide objective analysis tools. That's what users want, and it also makes life a little easier for Jiri, in that he did not have to care about how Rybka was developing
What is the relation between search and knowledge in Rybka 4?
Knowledge guides the search - Rybka examine those things that it's heuristics tells it to be important .
__________________________
By courtesy
ChessBase Magazine No. 135 / ChessBase GmbH
_________________
"No one is never ever truly gone as long as you remember them "
