This is part b of the interview, click here to see part a.
What is the secret of Chess Assistant's speed, and how do you manage to fit CAP, the 2 million game database, direct tree, opening book, and player encyclopedia all on two CDs?
Practically all the programmers that are on the Chess Assistant team are graduates of the MSU computer science department. Making their code powerful, quick and compact is foremost in their minds. So we never have had problems with such things. Nice visual design work is probably the most difficult for us.
You don't have a huge marketing effort behind Chess Assistant right now. How difficult has it been to get the word out about the program? Do you feel like much of your sales stem from word of mouth?
The chess database market is a very interesting market. A lot of people don't believe the ads they see, they believe their friends more. A case in point is CA3. Initial sales were good enough, then they went down. The sales of CA4 were not good from the start because a lot of people shared their opinion about CA3 However, with CA5, things started to turn around, and even grew. And CA6 was very interesting - sales were strong even after we released CA 7; i.e. most people were satisfied with the product and recommended it to their friends.
Historically, our marketing was never that intense. I think that a lack of money was
the main problem. In fact, I feel that it is serious flaw and that we should address in the next few years. It is especially important for teaching programs that have the same importance for our company as databases. Maybe database is not the best choice of words, since modern commercial databases (Chess Assistant, Chess Academy and Chess Base) are really integrated chess “offices” that provide almost everything that a chess player needs.
What areas of technology do you think will have the greatest influence on chess players in the coming years?
Proper synergy between standalone computers and network technologies will be one of the factors. Standalone computers are certainly powerful and fast, and people can do a lot of things using them. But if we can harness the power of a large number of computers, and quickly deliver the results to the user, it will give him/her much more
power. The main issue is intelligent use of network functions. Things like searching a five-man position in a network database doesn't make any sense to me.
Another thing is intelligence. Programs will have smarter algorithms, and the smartest programs will be on the top.
Do you think it is worthwhile to develop a Linux version of Chess Assistant? Are there any plans in that area?
Chess Assistant is very large program (more than 1 million lines of code), and adapting it to Linux is hard work, even with the Borland Kylix system (Kylix compiles Borland Delphi code to Linux). Debugging and tuning, all demand resources. Also, we don't think that a sufficient number of people will buy the Linux version. So making a Linux version is a luxury for us, but maybe this will change in the future.
What do you think is the greatest advance in the area of computer-aided analysis (i.e. tree, engine analysis, etc)?
There was a time when engine strength was not sufficient for analysis. But these days, engines are strong enough for good analysis of most games. Perfect tactical play, smart game analysis, generation of exercises - all of these are possible at the current stage. However, in some positions, engines still don't see simple strategic moves that even a 2000 ELO player can see. Plus, they are not able to teach ideas and plans like a human player can. These weaknesses, even though they occur maybe 10% of the time, hold back the utility of engine analysis functions.
So I think that the key is additional control of the engine thinking process. That is, some things can be improved with help of control programs that don't work with low-level position evaluation factors, but instead compare evaluations and variations of the analysis and provide the engines with new tasks. But some things can be improved only with the help of a human. So facilitating the display of critical points in the analysis process and having the ability to propose new variations is important.
Looking at these control functions is one of our top priorities.
I understand that Maxim Blokh is working for Convekta, what insights do you expect him to bring to the future development of your products?
There are a lot of good chess players and coaches. And it seems it is easy to write a program using their advice. But the reality is more complex. First of all, not many people can formulate their opinions clearly, and in terms understandable to the typical programmer. Secondly, it is not easy for a non-programmer to understand what is easy or difficult to do from a software standpoint.
GM Eugene Gleizerov continues to be a big help and provides many inputs, but we also need someone full-time at Convekta. So Maxim Blokh joined our team. Being a children’s coach and correspondence chess GM, Maxim knows well the needs of chess players of ALL levels, from 5 year-old kids to grandmasters. As I mentioned earlier, Maxim Blokh was instrumental in the design of CT-ART.
And I should say that the results are much better than we expected. Some results are visible today in CA7 but much more is coming. For instance, there a new test mode for creating and displaying exercises from analyzed games. Look also at the new infinite analysis mode. Only an experienced chess expert could come up with an idea like adding simultaneous analysis with the opposite side to move, to show threats. Or, coming up with the idea to ignore moves, or try only specified moves, all concurrently.
Was it hard to get Per-Ola Valfridsson to permit Ruffian to be bundled with CA? What do you think is the future of this very strong engine (Editor's note: Ruffian just won CCCT-5)?
Really it was not difficult to agree with Per-Ola, we found we had some common interests. And Dann Corbit was a big help with this as well.
Ruffian is a very interesting attacking program. I am still amazed how it is possible to create non-commercial programs with playing strength close to that of commercial engines. I found that in Sicilian games especially, Ruffian can outplay most commercial programs. I hope that new versions will be even more impressive.
This is part b of the interview,click here to see part a.