Back

  


How to Beat your Chess Computer


By Boris Alterman

As we know, the main problem of all chess programs is the very large number of continuations involved. The number of all potential positions is in the order of 10128 (10 to the power of 128), which is vastly larger than the number of atoms in the known universe (a pitiful 1080). It is clear that no computer or any other machine can solve the game by looking at all possible continuations. But we beings are also imperfect players. For the computer, it is only a question of what depth of search is required to match human strategic skill. The machine Deep Blue, which Kasparov played in Philadelphia and New York, consisted of an IBM SP/2 server equipped with a large number of special-purpose chips, which do the fast calculations. Each chip is capable of processing two to three million positions per second. By using more than 200 of these chips, the overall speed of the program could be raised to 200 million positions per second. In this connection we should remember that it was the chess chips that were generating the speed, not the host SP/2 machine, although IBM liked to suggest otherwise The conclusion of the experts: you need to build a computer that runs at one billion nodes per second (and searches 14 ply deep) if you wish to challenge the human World Champion for his title. Deep Blue comes close, but isn't there quite yet. The quality of programming also plays an important role. Today's top PC programs such as Deep Fritz and Deep Junior run at up to 1,000,000 positions per second on the fastest available hardware, and realistically have a playing strength in classical chess of over 2600. They are a match for all but the top 100 players in the world; in rapid forms of chess only the top dozen or so humans can compete. The extensive opening books that are used play an important role in the strength of computers. The collective knowledge and experience of many generations of human masters can easily be stored on a hard disk and accessed during the opening phase of the game. Even the micros know tens of millions of opening positions, and have access to full statistics for each of them (which moves were played, with what success, by what caliber of player, etc.). Very often a program will play fifteen or more moves of a game before it starts calculating moves for the first time. Without the benefit of this human knowledge in the openings, the programs would be considerably weaker. While computers are gaining a substantial advantage from the vast amount of opening knowledge that has accumulated in the history of chess, they also profit from research at the other end of the game. Once again it was the ubiquitous Ken Thompson who pioneered this development. In the 1980s he began to generate and store all legal endgame positions with four and five pieces on the board. A typical five-piece ending, such as king and two bishops vs. king and knight, contains 121 million possible positions. With a pawn, which is asymmetric in its movements, the number rises to 335 million. Thompson wrote programs that generated the positions and worked out all forcing lines that are possible in each endgame. He also compressed the resulting data in a way that allowed one to store about 20 endgames on a standard CD-ROM. Using these endgame databases a computer will play them absolutely perfectly ("like a god"). In any position with the given material on the board it knows instantly whether it is a win, draw or loss and how many moves it will take to reach that result. Often it will announce a win or mate in more than fifty moves. On the losing side it will defend optimally. Deep Blue was using Thompson's endgame databases in its search, but today the main PC chess programs have incorporated them into its search tree. When you have a match between such different opponents as humans and computers, you have to look at the different strengths and weaknesses of both sides during the competition. Playing against a chess computer means facing something that doesn't have any nerves - similar to sitting across the table from an IRS agent during a tax audit. It's quite clear what the weaknesses of a human being are - primarily our vulnerability to outside interference. We could catch a cold, be easily distracted, and so on. Obviously, we are not in the position to calculate as deeply as the machine, but we can compensate for that. It's less clear what the weaknesses of the machine are, but a computer specialist or a chess specialist can point them out. >At the top of any evaluation by a computer, we always see the material. It always tries to translate quality and time factors into numbers that represent the mathematical balance of material. As for the opening strategy, this is quite a dangerous part of the game now, since there is a huge amount of theory on specific openings and many lines that can be exploited further by the GM-team supporting the machine. My first principles in games against computers to avoid the main lines and to accept an inferior position after the opening, hoping that outside of the theoretical routes the machine will lose its horizons and will start making positional mistakes. The negative side of such a setup is that such a strategy dramatically limits your active opportunities; nonetheless, this decision has worked perfectly well many times. Computers often overestimate their chances and make positional mistakes, giving me a serious positional advantage. For example, Fritz may weaken its king without much hesitation, simply destroying the pawn protection of the king and not paying much attention to the king's safety before it becomes too late. Now we will see some examples of how to beat your computer:

Alterman,B – Computer Junior

Follow the moves with your own board
1.c4 I would like to start with a very illustrative game to demonstrate my first experience against the Computer Junior. It was played in the Nir Ganim tournament with a time control of 15 minutes +10 seconds per move. 1...Nf6 2.Nf3 c6 3.b3 d5 4.Bb2 First of all, it is quite good to escape from known opening theory. Computers have big problems when they leave the book immediately, after the first 3-4 moves. Computers have more problems making a decision when there are many pieces on the board. When a game leaves known theory, some of the pieces are often exchanged, giving the computer a much easier time.4...Bg4 5.e3 Nbd7 6.Be2 e6 7.h3 Bxf3?

Of course, this is a mistake. Junior doesn’t like to lose a tempo and therefore exchanged his bishop. Now White has a pair of bishops, and therefore has a slight advantage. Better is 7.Bh5. 8.Bxf3 Bd6 9.Be2 dxc4?

Another positional mistake. White dreams about opening up the position for the bishops, but the computer did so voluntarily!10.bxc4 0–0 11.0–0 Qc7 12.Nc3 Rfd8 13.Qc2 a6 14.Rac1 e5 15.Rfd1 White already has a quite clear positional advantage, but still not a winning position. I continue to play without taking risks or allowing tactical complications, and the computer eventually started to lose control.15...Nc5 16.d3 Ne6 17.Bf1 h5?

It works! My prophylactic quiet moves finally give the computer the feeling that he may activate his position on the kingside.18.Ne2 h4 19.d4 Finally after good preparation I start an action in the center. The computer created a clear target – the weak pawn on h4. I should launch an attack on the kingside.19...e4 20.d5 Nc5 Junior doesn't sense danger and allows me to destroy his kingside. Better was [20...cxd5 21.Bxf6 gxf6 22.Rxd5with a clear advantage for White.] 21.Bxf6 Bh2+ 22.Kh1 gxf6 23.Nd4 Rac8 24.dxc6 bxc6 25.Qe2

Finally, the white queen found her way to the kingside. White’s attack becomes unstoppable. 25...Be5 26.Qg4+ Kf8 27.Nf5 Ne6 28.Qxh4 Kg8 29.Qxe4 Kf8 30.Qh4 Kg8 31.c5 a5 32.Bc4 Another piece joins the game.32...Rb8 33.f4 Bb2 34.Rxd8+ Rxd8 35.Rf1 Rd2 36.Rf3 Rd1+ 37.Kh2 Rd5
And Black resigned due to the mate after Rg3 and Qh8 1–0