Chess x.x Chess Program
by Bill Wall
Before there was Vasik Rajlich and Rybka, before there was IBM and Deep Thought and Deep Blue, there was the Slate/Atkin program called Chess x.x.
Chess x.x (Chess 2.0 to 4.9) was the modified version of the chess program Chess, which was a computer program written in assembly language at Northwestern University (Vogelback Computing Center) in Evanston, Illinois by David J. Slate, Larry R. Atkin, and Keith Gorlen. The program later received support from Dr. David Cahlander of Control Data Corporation (CDC) in Minneapolis, a CDC Cyber hardware consultant.
In 1968, undergraduates Larry Atkin and Keith Gorlen decided to write a chess program to exercise Northwestern University’s new CDC 6400 mainframe computer ( a slower and cheaper version of the CDC 6600 supercomputer, the world’s fastest computer at the time – 3 million instructions per second). Upon hearing of their work, physics graduate student David Slate (rated around 2050) decided to write a competing program. They combined their programs in October 1969, and produced Chess 2.0.
In 1970, Slate received a letter containing suggestions for improving Chess 2.0 from International Master David Levy, who had tested the program at the University of London. Improvements were made and the new release was now called Chess 3.0. Chess 3.0 was now more efficient and running 65% faster than Chess 2.0.
Chess 3.0 through 4.9 was one of the most successful chess programs during the 1970s. Chess was the first published use of the bitboard data structure applied to the game of chess. Each bit represented a game position or state, designed for optimization of speed and/or memory or disk use in mass calculations. The first bit represents the a1 square and the 64th bit represents the h8 square.
In September 1970, Chess 3.0 won the first Association for Computing Machinery (ACM) U.S. Computer Chess Championship in New York. Chess 3.0 ran on a CDC 6400 and won all of its three games. There were 6 programs in the event.
Gorlen left Northwestern in 1970, but Slate and Atkin continued to work of their chess program.
In 1971, Chess 3.5 won the 2nd annual ACM event in Chicago, winning all of its three games. There were 8 programs in the event.
In 1972, Chess 3.6 won the 3rd ACM tournament in Boston, scoring 3-0. There were 8 programs.
In 1973, Slate and Atkin wrote a new program, Chess 4.0, rather than modifying the Chess 3.x series. A library of 5,000 opening positions was added.
In 1973, Chess 4.0 won the 4th ACM tournament in Atlanta. It won 3 games and drew 1 game. There were 12 programs.
A major turning point from Chess 3.0 to Chess 4.0 was the transition to full-width search and brute-force search to take advantage of the speed and computational capacity in the new computers.
In January 1974, Chess 4.0 played in a chess tournament with 50 humans at Northwestern University. It tied for 3rd place, scoring 4.5 out of 6. Its performance rating was 1736.
In 1974, Chess 4.0 participated in the first world computer chess championship in Stockholm, scoring 3 wins and 1 loss (losing to Chaos). The event was won the USSR program Kaissa. There were 13 programs in the event. After the event Chess 4.0 played one game against Kaissa, which was drawn.
In 1974, Chess 4.2 failed to win the 5th ACM U.S. Computer Chess Championship, held in San Diego. The University of Waterloo program RIBBIT (later call TREEFROG) won that event by defeating Chess 4.2.
On October 21, 1975, Chess 4.4 won the 6th ACM North American Computer Chess Championship (NACCC) event in Minneapolis with a perfect 4-0 score, using the faster CDC Cyber 175 computer (2.1 megaflops). There were 12 programs in the event. It was estimated that a doubling in computer speed increased playing strength by about 100 points.
In 1976, Slate and Atkin added a transposition table for Chess 4.5. Its rating was under 1600, or Class C level. After 10 years of development, chess programs gained less than 200 points. At that rate, it would take another 60 years before a computer could challenge the world chess champion. But in just a few years, Chess 4.9 would be playing at the Expert level.
On July 25, 1976, Chess 4.5 (rated 1579) won the Class B section of the 4th Paul Masson chess tournament in Saratoga, California with a perfect 5-0 score. It had a performance rating of 1950. This event was the first time any machine performed successfully in a tournament for humans and won a prize ($700, but was turned down by the programmers). After the event, Chess 4.5 was rated 1722.
In October 1976, Chess 4.5 won the 7th ACM NACCC tournament in Houston. Chess 4.5 was searching trees with 800,000 nodes per move using a CDC CYBER 176 (4.6 megaflops). It could look at 1,500 positions per second.
On February 20, 1977, Chess 4.5 won the 84th Minnesota Open Championship with 5 wins and 1 loss. It defeated one expert, Charles Fenner, rated 2016 and lost to Walter Morris, rated 2175. Its performance rating was 2271. Chess 4.5 then qualified for the Minnesota State Championship, but did not win it.
In March 1977, Chess 4.5 gave a simultaneous exhibition in New York, winning 8, drawing 1, and losing 1 (to Eric Bone – 2150). It then played 4 games of blitz chess against International Master David Levy, winning 2 and losing 2. Its blitz performance rating was 2300.
On August 9, 1977, Chess 4.6 won the second World Computer Chess Championship, held in Toronto. There were 16 participating programs from 8 countries, including defending champion Kaissa of the USSR. It won with a 4-0 score, defeating BCP, Master, Duchess, and Belle. 2nd place went to Duchess. In a special exhibition the next day, Chess 4.6 beat Kaissa.
In October 1977, Chess 4.6 tied for 1st place with Duchess at the 8th ACM North American Computer Chess Championship, held in Seattle. Both scored 3.5 points out of 4. The winning trophy was awarded to Chess 4.6 base on tie-breaking points.
In September 1977, Chess 4.6 achieved a 2000 rating in a tournament in London. On September 18, 1977 it was the first computer to beat a grandmaster when it defeated GM Michael Stean in London.
On April 30, 1978, Chess 4.6 won the Twin Cities Open in Minneapolis with a perfect 5-0 score. Going into the event, the program had a USCF rating of 1936. After the event, its rating was 2040.
On May 6, 1978, Chess 4.6 defeated U.S. chess champion Walter Browne (2560) at a 44-board simultaneous exhibition in Minneapolis. Chess 4.6 was running on a Control Data Corporation (CDC) Cyber 176 supercomputer and examining 2.5 million positions in three minutes of think time. Browne was the first grandmaster to lose a game of chess from a computer, but it was a simultaneous exhibition and not a normal tournament with time controls.
In August 1978, Chess 4.7 played a 6-game challenge match with David Levy (2350) for his famous bet in Toronto that no chess computer could beat him in a match by 1984 (he later won his bet). Chess 4.7 did not defeat Levy in the match, but it did beat him in game 4. Levy became the first International Master to lose a game to a computer in a tournament environment.
In October 1978, the 9th ACM NACCC tournament was held in Washington, DC. Belle won the event. Chess 4.7 came in 2nd place.
In October 30,1979, Chess 4.9 won the 10th ACM NACCC tournament in Detroit and was consistently playing at the expert level (2100).
In 1979, Chess 4.9 played in the 3rd world chess championship in Linz. There were 18 participating programs. Chess 4.0 score 2.4 out of 4 games. The event was won by Belle.
In 1980, Chess 4.9 drew a game with Larry D. Evans (2393) at the U.S. Amateur Team Championship. Chess 4.9 had a performance rating of 2168. In speed chess, Chess 4.9 performed at a 2300 Elo rating. Grandmaster Fridrik Olafsson had not trouble in defeating Chess 4.9 in two blitz games that year.
By the end of 1980, Chess 4.9 was retired from competition after Slate teamed up with William Blanchard to create NUCHESS. Atkin went to Applied Concepts and worked on dedicated chess computers such as the Great Game Machine and the Chafitz modular game system.
The Slate/Atkin program remained the best chess-playing program throughout the 1970s. It gained chess strength and rating with each new, faster generation of computer hardware. For every fivefold increase in computer speed, there was a 200 point increase in the program’s rating as it approached the master rating level of 2200.
References:
Atkinson, Chess and Machine Intuition
Botvinnik, Computers in Chess
Frey, Chess Skills in Man and Machine
Hsu, Behind Deep Blue
Levy, The Chess Computer Handbook
Levy, Computer Chess Compendium
Levy and Newborn, How Computers Play Chess
Marsland & Schaeffer, Computers, Chess and Cognition
Newbor, Computer Chess
Welsh, Computer Chess I and II
Welsh & Baczynskyj, Computer Chess II
Wall, ACM Computer Chess http://www.geocities.com/SiliconValley/Lab/7378/acm.htm
Wall, Computer Chess History http://www.geocities.com/siliconvalley/lab/7378/comphis.htm