Lovass László

Why Do Bitbiters Want to Eat Chess?

Reading or watching sci-fi classics we often face the final question: can our intelligent artifical creatures - called robots by Capek and Asimov - defeat us humans? Can we dominate them on their progress which is ten thousand times faster then the human evolution?

As chess is a kind of struggle, people often consider that losing on the chessboard is losing against a superior intelligence. Although it is not completely true - you may think of goddess Fortune's activity. But, let's say, winning against somebody 7 to 3 in ten games means that the winner can solve a special problem - finding a narrow path leading to victory - definitely better than the loser.

Computer specialists researching artificial intelligence are interested in studying chess programming. Why? At first, in many areas of human intellectual endeavour it is difficult to assess accomplishment accurately. But there is a statistically reliable rating scale in chess developed by Arpad Elo. The second reason is that the final aim - the checkmate - can be defined simply, but the way leading to that is extremely complicated. The third one is that in contrast with other types of games, the strategic rules don't always dominate the course of the game. Just think of some wonderful sacrifices for checkmating the opponent with only one or two pieces.

I think it is a kind of fashion to talk about artificial intelligence. But these days the so called artificial intelligence is rather 'artificial instinct': although those programs can solve difficult tasks, but they - at least, the most of them - can't change the basic methods created by the programmers. It seems to be a fact that it is easier to write a program that plays at ELO level 2400 than to make another one that knows only the chess rules, but by playing against a partner or itself can learn how to play chess, even if the program will never play stronger than ELO level 1900. Of course, the second program means the real artificial intelligence for me.

It's also obvious that mainly the `hardware dopping` is responsible for the muscular strength of the best chess programs these days. This is demonstrated in active chess (30') games against top players, where the speed often dominates over the correct way of playing.

You can hear totally different opinions on chess computers these days. But never forget that it could increase the popularity of chess. A chess computer might be sunshine in the stormy sky of a lonely person's life; can be used as a faithful and patient mind trainer in the old ages and even ready for a game late at night.

There are two challenges for chessplayers who face the power of computers. The first is a wide range of playing softwares or special chess machines, from the simplest ones you can put in your pocket to programs written to the most powerful personal computers with megabytes of memory and big storage capacity (harddisk, CD-ROM). But they are all constructed to find the possibly best move at any position.

The other kind was born by the 'big bang' of chess information. From the beginning of the chess history, the preparation before playing became more and more important: what kind of openings are played by the opponents, how to refine one's own repertory. As the quantity of the theoretical material increased, it seemed to be logical to use the faster and cheaper computers instead of Informators, bulletins and chess journals. Several database systems were developed.

Computer: Weapon in the Jungle of Information

About a decade ago some of my chessfriends proudly talked about their plans of advance: studying chess 5-6 hours a day, collecting all the important chess information for perfect preparation, tournaments, where they can show the fruit of their talent and diligence. I said that's all fine, but what do you mean by 'perfect preparation'? After a short discussion we agreed that in the world of mass-production you might be anxious that your opponent knows something important you have forgotten to review.

Think of the joke about the American and the Japanese tourists walking in the desert. Suddenly a hungry lion appears and starts to run towards them. The Japanese quietly stops and begins to change his shoes to a pair of running shoes.

'What are you doing?' the American asks him. 'Do you really think that your sports shoes make you run faster than the lion?'

'Of course, not' the Japanese replies, 'But at least faster than you!'

To avoid this, some good players worked on detailed card- index systems, but this endless paperwork made them rather theoretical specialists avoiding of the risk at the board. At dawn of home computing some of them tried to do this using those computers instead of playing with beginner level chess programs. When I asked some high rated players about this method, the most of them answered: oh, no, it is just a waste of time to learn how to use a complicated machine for this task. But machines became cheaper, faster and - which is perhaps the most important - more user friendly for (at least) the experienced users.

The first popular chess database called ChessBase came in 1987. The basic idea was to make a working card-index system - technically called database - on personal computer. This program worked using keys. A key contained information (e.g. English opening - Double Fianchetto) involving a set of games. Using these keys one after the other different openings could be found in this database. And, of course, player names and other informations could also be searched for. When I tried this program for the first time, I was slightly disappointed. It seemed to be the beginning of a brave new world, taking over the old-fashioned classification and opening code methods. But we make moves on the chessboard, not codes! Imagine that using a huge game database you can find thousands of games available even under a detailed code. This magnitude of games requires at least hours of additional selection.

How can we break through this bottleneck?

Being a chessplayer, my dream was an easily upgradable, multi-function tree-structure (likeEncyclopedia of chess openings). So I developed a new database system called SuperPro. Using this system you can follow move by move all the alternative lines, or switch between them. Contrary to Encyclopedia, these lines are games from the first move to the last. There can be also analysed lines and games being attached to each other like branches of a tree. For the first time, my program was written for small Sinclair Spectrum computer. It was very powerful even on that tiny system.

But SuperPro was designed rather to be an expert system than a simple database. What is the difference between database programs and expert systems?

Let's suppose you are a medical specialist in a hospital. You can not work without a database system gives you up-to-date information about the patients' diseases. You can search for symptoms, history of disease, medicines, patients' age, sex, occupation etc. Now, let's change it to an expert system created to help making diagnosis. At first, you can use the same database features. The real difference begins with the ability of asking the program's opinion. For example: you give the patient's symptoms, age, medical precedents etc. then the program suggests a possible diagnose (giving also the percent probability and background of its opinion) and may also suggest an effective treatment. This system would accelerate medical work and may show details which avoided even an experienced specialist's attention.

Well, it is not a fairy-tale, these computer systems work in many clinics and hospitals. Maybe many chessplayers don't believe in this kind of help of machines. But just think of the increasing number of games in database collections (magnitude of million games by now) and ask yourself: what can I do with this number of games? Is it too many or not? Do I really need them all? Any information can be useful for personal preparation. If a psychologist can evaluate personality from many small signs (e.g. tests), why we, chessplayers, would ignore the possibility of getting known somebody from his/her games? Even, computer assistance may discover details you'd never suspected!

Being disturbed by the perpetual stream of games, think about the saying of Mr Albert Szent-Gyorgyi, the Nobel prized Hungarian scientist: 'The books are for keeping the knowledge while we should use our minds for something valuable - thinking.' is also true for database systems. Analysed material with basic games selected by well known theorists would solve this problem (at least, in the case of the popular lines) - it is a part of the chess background of SuperPro Chess to issue such collections. And your own research can make a sharp sword of surprise for your opponent from an opening sideline.

As mentioned above, using SuperPro you can study all possible alternative lines (games) move by move and even add your analysis to any line you want. Don't forget, Dear Reader, that using a chess set for this you spend about 20-25 percents of time resetting the starting position at each alternative line you want to examine (at least 8-10 months in a chess fan's life!). It's even more difficult to estimate how much time is wasted on sharing concentration between the board and the printed analyses to follow.

In SuperPro not only the board, the starting and current moves of the game, the alternatives, the game header and the material balance of the endposition but also the statistical information about branches can be seen as you move from one to the other. For this you never have to concentrate on opening keys or inputting/ importing double games by accident. New games immediately find their own places so it's impossible to enter the same game again. Therefore a database of 1 000 000 games consumes only about 84 Mbytes.

You can search and reorder tree by any kind of information: players (both for white and/or black, free of accent problems), tournaments, years, results, move number intervals, openings, positions, move sequences, endgames, pawn structures, piece positions and the combinations of these, also with the use of repetory filters.

SuperPro can use prepared (in Dbase or text format) lists and databases (e.g. ELO list). Selected name list can be created for mass-search (84 names at once or separately). If this list is about possible tournament participants, ELO calculator will work out ELO average and expectation.

Imports and collections can be made from other database formats (CBF, PGN) or text files. Having phone and modem or serial link cable, two SuperPros (on two PCs) can communicate with each other for playing chess, remote analysis and search.

The Test feature is perfect for chess lessons: you can find out the following moves after a diagram position with the desired time limit setting and recording your score. Either the prepared test database or the games of a new tournament after marking diagrams (e.g. by sac search) can be used for this purpose.

SuperPro uses Genius chess program to examine all marked positions and games then transfers these analyses into the database. Using this feature at night, you might get the chess program's opinion about all critical positions of your repertory for breakfast!

These are just examples from the program's abilities. If you are not satisfied with all possibilities you'll be able to make your own commands (macro). This way learning chess is fun!

August 1999

(Up-to-date version of the original article published in the complimentary issue of Black Is OK Magazine)