Computer chess

from Wikipedia, the free encyclopedia
Tandy RadioShack 1650 from the 1980s

Computer chess or computer chess denotes the playing of chess against a computer , the playing of computers among themselves, the development of chess-playing machines ( chess computers ) and the development of chess programs .

The idea of ​​creating a chess-playing machine dates back to the 18th century. For more on the history see under chess computer , on how it works under chess program .

Computer chess as a research subject

The main goals of computer chess were entertainment, chess analysis, and the hope for insights into human thinking. Since the mid-1960s, computer chess has often been referred to as the “ Drosophila of Artificial Intelligence ”. While the first two goals were achieved with flying colors within 50 years, the hope for insights into human thinking was disappointed. All research in this direction (e.g. by Michael Botwinnik or Allen Newell ) was never crowned with success.

For this reason, computer chess (like Scrabble ) was no longer a subject of research for a long time and was largely replaced by games such as Go or Arimaa , since computer programs can achieve success in these less through pure computing power than through a much more complex evaluation function and both games are still relatively easy People can be learned and played successfully.

Instead, the ongoing miniaturization and often doubling of the computing speed of computers has played into the hands of the brute force advocates of chess: chess computers for household use are nowadays available at negligible costs and there are a number of chess programs ( open source and freeware programs such as Fruit , Amy, Pepito, Crafty and others), which are on a par with grandmasters on standard PCs . Top programs such as Shredder , Junior or Fritz even beat the world's best in tournament thinking times on a regular basis.

It remains to be seen whether the computers will make the game of chess uninteresting in the foreseeable future, as their skill levels are constantly increasing. However, it is argued that even with computer programs getting better and better, the game of chess would remain interesting - after all, people would still compete in sprints or marathons, even though every motorized vehicle is faster. One possibility is to only use strong enough hardware in show fights that the software is roughly on par with the opponent. While Deep Blue was able to analyze around 200 million positions per second on specialized hardware in 1997, Deep Fritz was only able to analyze around eight to ten million positions per second on commercially available hardware in 2006. However, the hardware's speed disadvantage has been compensated for by the software's improved sorting, search and evaluation algorithms.

In 2017 AlphaZero stunned the chess world. The Google researchers of the Alpha Zero group published 10 games from a tournament of 100 against Stockfish  8 , which AlphaZero won with 28 wins and 72 draws without defeat. The program had taught itself to play chess in just a few hours, playing millions of games against itself, just knowing the rules of the game. The project was and remained otherwise in a quiet room.

Shortly afterwards, an open source project of a similar kind was started: Leela Chess Zero (LC0). This runs on normal PC hardware. At the end of 2018, LC0 achieved a sensational result in the world computer chess tournament, where it defeated all current chess programs until it was narrowly defeated in the decisive battle.

One of the interesting things about this new development is that these AI chess programs are amazingly “human”. Position play and strategy come into play here much more. With LC0 you can adjust the playing strength to your own abilities by selecting older network files. Then LC0 actually just plays weaker - like a human. With conventional chess programs, on the other hand, one has the impression that the otherwise perfectly playing machine simply throws in a bad move every now and then.

Strategies against computers

While humans can make longer-term plans and occasionally overlook short-term threats, computers exploit every tiny tactical flaw. The programmers try to teach their programs better and better strategic “knowledge”. However, there are particular problems with how to evaluate a position. Roughly speaking, a chess program tries every possible move (and all possible moves on it up to a certain depth) and evaluates the resulting positions using an evaluation function. Many positions are difficult to evaluate with a number. Features like pawn structure, open lines etc. often have advantages and disadvantages for both sides. People who lack the computing power of a computer cannot play through every move in their head and consider the resulting positions. Rather, over time, a feeling ( intuition ) emerges as to which move in which position could result in an advantage. These moves are then examined more closely.

Computers are clearly superior to humans in tactical maneuvers that can be completed within their computing depth. The queen is particularly dangerous , so that human players often try to persuade the computer to swap queens. It is in the nature of things that such “tricks” - once recognized - are taken into account by the programmers in subsequent versions when programming.

Strategically, a person has to operate against a computer with long-term maneuvers, the approach of which is initially not recognizable for the computer within the scope of its computing depth. So had z. B. Wladimir Kramnik once against Deep Fritz Success with a long-term possible march through by a passed pawn who - initially not yet drawn far - was recognized by Deep Fritz too late as a serious threat. The computer thus punishes combinatorial strategies and enforces a positional play system.

Another strategy is to play an unusual move at the beginning to quickly force the computer out of its opening repertoire. So he has to start calculating the cheap moves in a time-consuming way instead of looking up them in a table. However, this also harbors risks for the human player.

Chess computer in the game against humans

The Levy Bet

In 1968 the Scottish International Master David Levy bet with several computer scientists for 1250 English pounds that within the next ten years no computer program would manage to defeat him in a competition. In August 1978 there was a match in Toronto against the then best program Chess 4.7, which Levy was able to win with 3.5: 1.5. In 1979 he played a show game against an improved version of this program, which was broadcast on ZDF and ended in a draw. There was a second bet that ran for another ten years. In 1988 Levy had no chance against the Deep Thought program and lost 4-0.

AEGON tournaments

Between 1986 and 1997, tournaments between chess computers and human players were held annually in The Hague, financed by the insurance company AEGON. In the first few years only Dutch amateurs played against the computers, later well-known grandmasters such as David Bronstein , Jeroen Piket , Vlastimil Hort , John Nunn , Larry Christiansen and Yasser Seirawan were invited to stand up to the ever-improving computers. Although one person won the individual ranking in all twelve tournaments, the computers won the overall ranking for the first time in 1993.

Competitions against the world's best

Since the 1990s, chess computers have also become serious opponents for players of the world's elite, but initially only in blitz and rapid chess . On August 31, 1994 there was a sensation when the world champion Garry Kasparov lost 0.5: 1.5 at a rapid tournament in London against the Chess Genius program running on a Pentium .

The specially developed chess machine Deep Blue from IBM defeated Kasparow in 1997 in a media-effective competition over six games with a long reflection period. Since this version of Deep Blue only played these six games in public, not much is known about the level of play achieved. After the competition, Kasparov suspected that the victory of the machine in the second game was achieved with human help. No evidence was found to support this claim.

In 2002 and 2003 newer programs held a draw in show competitions against two of the world's best grandmasters ( Brains in Bahrain 2002 Deep Fritz against Wladimir Kramnik , 2003 Junior and again Deep Fritz against Kasparov).

In 2005 the grandmaster Michael Adams (in July 2005 number 7 in the world chess rankings) competed against the computer Hydra . The computing power of the program was approx. 200 million positions per second. The computer decided the match clearly with five wins and one draw. In the tournament, the machine remains unbeaten by humans. In correspondence chess, however , Hydra has already suffered two defeats against correspondence chess grandmaster Arno Nickel .

From November 25 to December 5, 2006, Wladimir Kramnik played a six-game match against Deep Fritz 10 in Bonn. Some conditions that were advantageous for humans applied: Kramnik received the program version used in the competition in advance in order to familiarize himself with their style of play. During the games he was shown the moves stored in the program's opening book . After 56 moves he had the right to a stalemate to apply, and he would have in positions of the final database will be rated accordingly can claim a draw. However, this did not happen in the competition. Deep Fritz won 4: 2 (2 wins, 4 draws).

In 2014 and 2016, the American grandmaster Hikaru Nakamura (best Elo number 2816 in October 2015) tried in vain to save “the honor of humanity”. Despite not insignificant requirements, he was defeated by the chess engine Stockfish in 2014 with 1: 3 and in 2016 by the engine Komodo with 1.5: 2.5 without winning the game. Sam Copeland compared the "brave" but futile resistance to Komodo with the historic battles of the Alamo and Thermopylae.

After the defeat of Kasparov, even the best human players began to train systematically with the help of chess computers. In particular, attempts were made to develop special strategies for competitions against computers. Nowadays people are already inferior to chess programs on a cell phone. To make matters worse, computers play their best chess with every move throughout the game, since, unlike humans, they do not tire and they do not make any obvious mistakes .

Competitions between chess computers

There have been and are a number of competitions and tournaments between chess computers. The North American Computer Chess Championship ( NACCC) and the Microcomputer Chess World Championship (WMCCC) were of international importance . The most important computer chess tournament is now the annual World Chess Championship (WCCC). The 25th WCCC took place in Macau in mid-August 2019 . As in the previous year, Komodo became world champion .

Another competition in computer chess is the Top Chess Engine Championship (TCEC), in which free chess programs also take part. The TCEC is seen by some as the unofficial world championship in computer chess. The most successful participant in the TCEC so far is the Stockfish chess program .

Computer chess in the media

Chess-playing computers are often motifs in films, for example in 2001: A Space Odyssey . In the series Starship Enterprise , the first officer ( Mr. Spock ) of the starship notices a malfunction of the computer when it loses in check against him and does not, as would have been expected, a draw against the Vulcan. In one of the Star Trek movies ( Star Trek IV ) , a computer-posed chess problem in 3D chess is one of the tasks by means of which Spock's mental abilities are checked after an accident.

literature

See also

Individual evidence

  1. Nathan Ensmenger: Is chess the drosophila of artificial intelligence? A social history of an algorithm. In: Social Studies of Science. Pp. 5-30, Vol. 42, No. 1, 2012, doi : 10.1177 / 0306312711424596 .
  2. D. Silver et al .: Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm. (PDF; 630 kB). S 5.
  3. Sam Copeland: Komodo Beats Nakamura In Final Battle. At: chess.com. January 8, 2016, accessed March 12, 2017.
  4. Breakthrough performance by Pocket Fritz 4 in Buenos Aires. At: Chessbase.com. Mobile version wins tournament in Buenos Aires (2009).
  5. Komodo 13 is world champion in computer chess. August 22, 2019, accessed August 5, 2020 .
  6. ^ Andrew Soltis : Engine Super Bowl. New York Post, accessed January 27, 2020 .

Web links

Commons : Computer chess  - collection of pictures, videos and audio files