Human vs. Machine

How computer chess went from a theoretical curiosity to completely superhuman β€” and changed how every player on this site's Champions pages actually studies the game.

The Theoretical Beginning

Long before any computer could actually run one, mathematician Alan Turing designed a complete chess-playing algorithm on paper in 1948-1950, evaluating moves by hand according to his own rules since no machine of the era was powerful enough to execute it. It's widely considered the first serious chess engine ever designed, even though it only ever "ran" as Turing himself manually following his own written instructions.

Early Engines Catch Up

Computer chess programs improved steadily through the 1960s-80s as hardware caught up with theory, moving from barely beating beginners to challenging club players, then masters. Dedicated chess computers became commercially available for home use by the 1980s, and by the early 1990s the strongest programs were competitive with strong grandmasters in specific games, even if not yet consistently dominant.

Deep Blue vs. Kasparov

IBM's purpose-built chess computer, Deep Blue, played World Champion Garry Kasparov in two landmark matches. Kasparov won the first, in 1996. In the 1997 rematch, an upgraded Deep Blue defeated Kasparov β€” the first time a reigning World Champion had lost a full match to a computer under standard time controls, and a moment that made headlines well beyond the chess world as a milestone for artificial intelligence generally.

The Engines Pull Away for Good

Within another decade, dedicated chess engines running on ordinary consumer hardware β€” far less exotic than Deep Blue's purpose-built system β€” surpassed every human player in the world, and have never looked back. By the mid-2000s, the question of whether a computer could beat the World Champion had been completely settled; the far more interesting question became how much stronger engines could still get, and what humans could learn from them.

AlphaZero Changes the Approach Itself

In 2017, DeepMind's AlphaZero was given nothing but the rules of chess and told to learn entirely through self-play β€” no opening books, no endgame tables, no human game data at all. Within hours of training, it was defeating Stockfish, at the time the world's strongest traditional engine, playing a strikingly different, more intuitive and sacrificial style than classical engines had ever shown. AlphaZero's success directly inspired the "NNUE" neural-network evaluation techniques that Stockfish and other traditional engines have since adopted themselves, blending the old brute-force search approach with the new learned-intuition style.

What This Means for Human Chess Today

Every elite player covered on this site's Champions pages today trains extensively with engines β€” not to compete against them (that contest ended decisively years ago), but to analyze positions, test opening ideas, and check calculations no human could verify alone. Engines are now simply part of the furniture of top-level chess preparation, the same way a calculator is part of doing advanced mathematics.

A handful of niche "centaur chess" events, pairing a human and an engine together as one team against other human-engine pairs, still explore what the best possible combination of human judgment and machine calculation can achieve together β€” though the era of humans meaningfully competing against engines alone, unaided, is permanently over.

An unhandled error has occurred. Reload πŸ—™