HomeLearn to Play ChessImprove Your GameChess HistoryChess for FunChess Blog

The Development of Computer Chess
The development of chess playing computers has paralleled the development of general computing.


Introduction: (August 2006) Since the dawn of the computer age, the development of chess playing computers has paralleled the development of general computing. Early computer pioneers and artificial intelligence (AI) researchers took an interest in chess because its fantastically large number of possible positions could not be tackled by brute force. The development of stronger chess playing programs has been a function of two types of technological advance -- better algorithms and faster hardware.

Historical Curiosities - The First Chess Playing Machines (au-'tom-a-ta): The first chess machine was the Turk, unveiled in 1769 by Wolfgang von Kempelen, at the court of Empress Maria Theresa. It was an early example of a cabinet illusion; a magician convinced spectators that no one was inside. A strong chess player, well concealed, made moves by manipulating the Turk's arm. Later examples were Ajeeb (1868) and Mephisto (1878). • The first real machine player was El Ajedrecista, unveiled by Leonardo Torres y Quevedo in 1914. It played King & Rook vs. Rook endgames.

Early Computer Pioneers - Alan Turing (1912-1954) and Claude Shannon (1916-2001): Turing, a British mathematician and the 'father of computer science', worked at Bletchley Park, Britain's codebreaking center during World War II. A keen chess player, he wrote a chess program before computers existed and stepped through a simulated game. Shannon, an American mathematician and the 'father of information theory', wrote a paper 'Programming a computer for playing chess' (1950). He also calculated the 'Shannon number', an estimate for the computational complexity of chess.

Algorithms and Data Structures - Start with Minimax, Alpha-Beta, and Trees: Although computers follow the same rules of chess as people, they can't use the same thought processes to select a move. They build a tree of possible variations, prune the tree, and follow the branch that seems to offer the best results. Shannon imagined two strategies. His Type-A strategy would build the tree to a fixed depth, then search for the best move using a procedure called minimax. His Type-B strategy would use chess knowledge to expand the tree for interesting variations.

Von Neumann (1903-1957), Herbert Simon (1916-2001), Allen Newell (1927-1992): In 1945 John von Neumann proposed the stored-program computer architecture and the first working machines appeared a few years later. Chess programs continued to receive attention as a testbed for the simulation of complex processes. Early AI work between the Carnegie Institute (later Carnegie Mellon) and the RAND Corporation, brought together Simon, Newell, and J.C. (Clifford) Shaw. They published 'Chess-Playing Programs and the Problem of Complexity' (1958) and developed the NSS chess program.

The 1960s - Software Engineers Program General Purpose Computers: In 1961-62 Alan Kotok, an undergraduate, M.I.T. wrote a Type-B program under the guidance of AI guru John McCarthy. Kotok lost interest, but McCarthy took the program to Stanford and improved it to play better than a rank beginner. In 1966, it lost a four game match to a Type-A program developed in Moscow. In 1966-67 Richard Greenblatt of M.I.T wrote Mac Hack VI. It played in several Massachusetts tournaments sanctioned by the USCF, ultimately achieving a performance rating of 1640.

The 1970s - The First Computer Tournaments, Microcomputers, and Custom Chips: In 1968 Scotsman David Levy (IM 1969) made a landmark bet with McCarthy and other AI experts that no program would beat him within ten years. He was not defeated until 1989. The first tournament for chess programs was held by the Association for Computing Machinery (ACM) in 1970, and the first World Championship, won by the Soviet program KAISSA, in 1974. The year 1977 saw the founding of the ICCA and the first computer with custom chips: Belle, developed by Ken Thompson of Bell Labs.

The 1980s - Hardware Engineers Take Control of the Playing Field: The computer chess technology advanced rapidly in the 1980s. The first microcomputer based machine, Chess Challenger, had been marketed in 1977, and the first World Microcomputer Championship was held in 1980. The machines gained their first win over a master (1981), their first master rating (1983), and their first win over a grandmaster (1988). At the U.S. Open in 1988, Deep Thought achieved a performance rating of 2745 and shared first place with GM Tony Miles.

ChipTest / Deep Thought / Deep Blue: Deep Thought was the evolution of work started in 1974 by Hans Berliner (World Correspondence Champion 1968-72) at Carnegie Mellon. His program Hitech tied for first at the 5th World Computer Chess Championship (WCCC), Cologne 1986, losing the title on tiebreak. Berliner hooked the attention of chip designer Feng-hsiung Hsu. Together with Murray Campbell, Hsu developed a machine called ChipTest (1985), which became Deep Thought (1988). Hsu and Campbell joined IBM in 1989.

Kasparov - IBM matches: In 1989, Deep Thought won the 6th WCCC at Edmonton but lost both games of a short match with World Champion Garry Kasparov. In 1993, the machine was renamed Deep Blue, a reference to IBM's nickname Big Blue.

Under ACM sponsorship, to mark the 50th anniversary of the first computer, the IBM team squared off against Kasparov, February 1996, in Philadelphia. Kasparov lost the first game, but won the last two to win the match +3-1=2.

A little more than a year later, May 1997, Kasparov accepted a direct challenge from IBM to play a match in New York. After winning the first game, Kasparov lost the second, resigning in a drawn position. He then drew three straight games before losing the last game in one of the most crushing defeats he had experienced in his career. By a score of +2-1=3, Deep Blue and the IBM team had accomplished the long sought goal of computer chess, beating the reigning World Champion in a set match.

Kasparov took the loss badly. He accused Deep Blue of having had human assistance during the games and demanded to see the software logs. IBM estimated that the corporation received 50.000.000 US$ worth of publicity during the match. A few months later the company announced that they were scrapping the Deep Blue project to tackle other AI challenges.

Man - Machine events: Since 1997, the advantage has tilted increasingly in favor of the machines. In 2002, Vladimir Kramnik drew a match with Fritz in Bahrain (+2-2=2). A few months later Kasparov drew with Junior in New York (+1-1=4). In 2004, three top GMs, Veselin Topalov, Ruslan Ponomariov, and Sergey Karjakin played a team match vs. Hydra, Fritz, and Deep Junior, at Bilbao, Spain. The match ended in a 3.5-8.5 rout for the humans, who could manage only one win and five draws. In 2005, world no.7 Michael Adams lost by a score of +0-5=1 to Hydra at the Wembley Centre in London.

The time may be near when humans are forced to accept material odds, like Pawn and move, in order to have a fighting chance.