🕰️ A Chronology of Computer Chess

Chess has always been more than a game. It has served as a laboratory for exploring the limits of intelligence—both human and artificial. What follows is a chronology of this remarkable journey, from the first mechanical device that could play simple endgames to today's LLMs, now omnipresent in our daily lives.

 

1910s – Leonardo Torres Quevedo's El Ajedrecista

In 1912, Spanish engineer Leonardo Torres Quevedo amazed the world with El Ajedrecista ("The Chess Player"), the first chess automaton. This machine could flawlessly deliver the king-and-rook checkmate against a lone king. It didn't ‘think' like modern engines, but it was a groundbreaking glimpse of the future of chess and technology.

 

1950 – Alan Turing's Turochamp

In 1948, Alan Turing and David Champernowne wrote the first chess algorithm, a set of instructions that would enable a future machine to play a game of chess. There were no computers yet, at least not capable of following complex instructions like the ones envisioned by Turing, so he tested it by hand, calculating moves step by step.

"This was a milestone: the idea that a machine could play chess existed decades before the hardware could make it a reality", explains Frederic Friedel. The program, named Turochamp, was the first of the so-called "paper machines".

 

1956 – The first computers

At Los Alamos and IBM, a team led by mathematician and engineer Alex Bernstein developed one of the earliest chess programs. It could only play on a simplified 6x6 board, but it marked a bold first step.

In 1957, the program ran on an IBM 704 mainframe computer. This giant, vacuum-tube machine weighed 9.7 tons, and the operator had to use levers to provide input and interact with it. It could process 42,000 instructions per second and had a memory of 70 kilobytes. Each move took a painstaking eight minutes to compute.

Bernstein's program followed Claude Shannon's "Type B" approach, applying criteria to prune brute-force search. With this, the seeds of computer chess as a serious research field were planted.

 

1960s–1970s – The start of the artificial intelligence race

University labs such as MIT, Carnegie Mellon, and Stanford refined early chess algorithms. Search methods like minimax and alpha–beta pruning became central, and by 1967, programs like MacHack could finally beat human players (at least beginners).

Computers were no longer just curiosities; they were starting to compete. Around this same time, American computer scientist John McCarthy coined the term "artificial intelligence," marking a new era in both chess and technology.

 

1966–67 – The Cold War in computer chess

When AI pioneer John McCarthy visited Moscow, he discovered that Soviet researchers were also developing chess programs. A challenge was soon issued: a game to be played via telegraph between an American-made and a Soviet-made chess engine.

The Soviet program, with former World Champion Mikhail Botvinnik as an advisor, won the match. It was later developed further and given the name Kaissa, the program that went on to become the first Computer World Chess Champion in 1974.

 

1977–78 – The first serious Human vs. Machine matches

Once again, a trip by John McCarthy sparked a chess challenge. During his 1968 visit to Edinburgh, he met reigning Scottish champion David Levy, a vocal skeptic of chess computers. "I'm sure that in ten years there will be a computer program capable of beating you!" McCarthy declared. A $2,500 bet was made, and the challenge was on.

Between spring 1977 and September 1978, Levy played a series of matches against several chess programs, none of which proved a serious threat. Then, shortly before the ten-year deadline, CHESS 4.7 (created by Larry Atkin, Keith Gorlen, and David Slate at Northwestern University) faced Levy in a six-game match. The program lost overall but managed to score one win and one draw, marking the first time a computer defeated a human master in a serious, competitive setting.

 

1980s – The rise of commercial software & ChessBase

With personal computers spreading rapidly in homes and businesses, fueled by MS-DOS and advancing hardware, commercial demand gave a strong push to the development of chess software.

In 1986, ChessBase was founded in Germany, revolutionizing how players prepared. For the first time, thousands of games could be stored, searched, and studied in a digital database. From amateurs to world champions, no serious player could afford to ignore computer databases again.

 

1985 – Kasparov humiliates chess programs

In June 1985, before he even became World Champion, Garry Kasparov gave a simultaneous exhibition in Hamburg, Germany. What made it unique was that all 32 of his opponents were chess microcomputers, developed by brands like Novag and SciSys.

Kasparov showed no mercy, winning every single game. Final score: 32–0 for humankind.

1988 – The first Grandmaster bites the dust

IBM's Deep Thought, an early precursor to Deep Blue, competed in an open tournament in Long Beach, California. There it caused a sensation by defeating Danish Grandmaster Bent Larsen, who had been one of the world's top players .

This was the first recorded victory of a chess-playing machine over a grandmaster — and a very convincing one, marking a major milestone. Now everybody was wondering: who would be next, and how long before the World Champion would fall? Kasparov still didn't see it coming: "Yes, Bent Larsen is a strong grandmaster. But a computer will never beat Karpov or me," he said at the end of 1988.

 

1996 – Deep Blue stuns Kasparov

The long-awaited showdown between the World Champion and a supercomputer finally took place in February 1996, when Garry Kasparov faced IBM's Deep Blue in a six-game match in Philadelphia.

The clash began with a shock: Deep Blue, a monster by 1990s standards, equipped with 256 specialized chess chips capable of evaluating 200 million positions per second, won the very first game.

This opening triumph sent shockwaves around the world, sparking fdebates about whether AI could not only rival but even surpass human intellect. Kasparov, however, adapted quickly, winning three games and drawing two to secure a 4–2 victory, keeping human pride intact, but leaving the world eager for a rematch.

 

1997 – Deep Blue beats Kasparov in a "best of six" match

In May 1997, Garry Kasparov returned to face IBM's Deep Blue in New York, in what became a historic turning point for chess and artificial intelligence. Unlike its earlier version, this upgraded machine featured faster processors, enhanced parallel computing, and refined evaluation functions, allowing it to analyze more than 200 million positions per second with improved accuracy.

The system was developed by a team of IBM scientists led by Feng-hsiung Hsu, Murray Campbell, and Joseph Hoane. After six tense games, Deep Blue triumphed 3½–2½. The globally televised event stunned audiences, and marked the first time a reigning World Champion lost a match to a computer.

 

1998 – Kasparov's "Advanced Chess"

Following his defeat against Deep Blue, Kasparov proposed the Advanced Chess format: humans and computers working together as partners. The city of Leon (Spain) was the first to held a match between Kasparov and Topalov under this format, which tried to prove the point that the strongest combination was not a computer alone or a human alone, but a human guiding a computer. It was a profound insight into "human–machine collaboration" that resonates with today's AI debates.

 

2000s – The engine era

Programs like Fritz, Junior, Shredder, and later Rybka came to dominate the chess world. Using brute-force search combined with evaluation functions, they grew far stronger than any human player. The real competition shifted from man vs. machine to engine vs. engine, as they climbed to ever-higher ratings. Around the globe, players began relying on engines for analysis and preparation, blurring the line between indispensable training tools and unbeatable rivals.

 

2010s – Stockfish and Open Source dominance

Stockfish, an open-source engine, rose to the top, consistently ranked as the strongest engine in the world. It set a standard for transparency and collaboration in engine development. Its evaluation function combined hand-crafted heuristics with ever-deeper search.

 

2017 – AlphaZero: A paradigm shift

DeepMind introduced AlphaZero, which learned chess entirely by self-play with no human guidance. Within hours, it surpassed Stockfish. Its style stunned the chess world: sacrificial, dynamic, and deeply creative. For the first time, engines weren't just brute-force calculators—they appeared to discover beauty.

 

2020s – Modern AI models and LLMs

Leela Chess Zero (Lc0), inspired by AlphaZero, introduced neural-network chess to the open-source community and transformed how players evaluated strategy. Its insights reshaped human understanding of positional play.

At the same time, modern AI language models (such as GPT-5) entered the chess world, with some decent attempts to explain, teach, annotate games, and generate educational content. While still outclassed by engines like Stockfish, these models have enriched the global chess ecosystem in entirely new ways.

In August 2025, Kaggle Game Arena, partnered with Google DeepMind, hosted a three-day AI Exhibition Chess Tournament, featuring eight leading large language models, including OpenAI's o3, xAI's Grok 4, and Google's Gemini 2.5 Pro. Held in a single-elimination format, o3 dominated, defeating Grok 4 in the final with a 4–0 score. Streamed live with commentary by chess grandmasters like Hikaru Nakamura, the event showcased AI's strategic reasoning, offering insights into their general problem-solving capabilities.