Deep Blue Defeats Kasparov, May 11, 1997: Match and Meaning

IBM’s Deep Blue defeated Garry Kasparov on May 11, 1997, winning the sixth game and the match by 3½ points to 2½. It was the first time a computer had beaten a reigning world chess champion in a match played under standard tournament time controls. The result became a durable symbol of machine intelligence, although the system was a highly specialized combination of custom hardware, search software and human chess knowledge.

Why the 1997 Match Was a Rematch

Kasparov and an earlier version of Deep Blue had played a six-game match in Philadelphia in February 1996. The computer won the first game, itself an important first under standard controls, but Kasparov adjusted and won the match 4–2 with three victories, two draws and one loss.

IBM’s team spent the next year rebuilding. Engineers increased processing power, improved the evaluation function that judged positions, expanded endgame resources and worked with grandmasters to refine opening preparation. The machine was not simply running the same program faster. Its designers were changing how it selected promising variations and assessed strategic features.

Kasparov, born in the Soviet Union and world champion since 1985, had made his reputation through deep preparation, tactical calculation and psychological pressure. A match rather than a single exhibition gave him time to study the opponent’s habits and adapt. That was why the 1996 recovery mattered—and why IBM’s 1997 challenge drew attention well beyond regular chess audiences.

The rematch took place at the Equitable Center in New York from May 3 through May 11. The dates are visible on the 1997 calendar, with the deciding game falling on a Sunday.

How the Six Games Reached a Tie

Kasparov won Game 1. Deep Blue answered in Game 2, a complicated contest that changed the emotional tone of the match. Kasparov later questioned one of the computer’s moves and wondered whether a human had intervened, an allegation IBM denied. The move was not proof of outside control, but Kasparov’s uncertainty became part of the match narrative.

Games 3, 4 and 5 ended in draws. That sequence left the score tied 2½–2½, making Game 6 decisive. Kasparov had Black. He chose the Caro-Kann Defense, a system he had used successfully, but allowed a known piece sacrifice that gave White a dangerous attack.

Deep Blue’s ninth move sent a knight into e6. Kasparov captured it, and the computer used the opened lines to keep his king under pressure. After only nineteen moves, Kasparov resigned. The game lasted about an hour, an abrupt ending to an event promoted as a prolonged test of human and machine calculation.

The final score did not mean the machine won every game or dominated throughout. One win apiece and three draws had produced the tie before Game 6. The historical claim is specific: Deep Blue became the first computer system to win a match against the reigning champion under standard conditions.

What Was Inside Deep Blue

The 1997 Deep Blue used an IBM RS/6000 SP system with 32 processors and 480 specialized chess chips. IBM says it could examine as many as 200 million positions each second. Speed allowed the machine to look through large trees of possible moves, but raw calculation needed guidance.

An evaluation function assigned values to features such as material, king safety, pawn structure and piece activity. Search extensions allowed the system to investigate tactically important lines more deeply. Opening databases provided established theory, while endgame information helped with positions containing fewer pieces.

Grandmasters worked with the development team. Their knowledge shaped evaluation settings and preparation, much as a human player’s team helps analyze openings. During play, tournament rules controlled interaction with the system; operators entered Kasparov’s moves and reported Deep Blue’s selected responses.

Calling the computer “brute force” is therefore incomplete. It searched vastly more positions than a person could, but people designed the scoring rules, hardware and preparation. Calling it human-like intelligence is also misleading. Deep Blue did one bounded task extraordinarily well and had no general understanding outside chess.

The Dispute After Game 6

Kasparov was visibly frustrated and called for a further match. He wanted access to complete machine logs and criticized IBM’s handling of information about earlier games. IBM retired Deep Blue rather than arrange another contest. Parts of the system later entered museum collections.

The refusal encouraged speculation, but the match result remained official. Later accounts by developers explained the design in technical detail, including an IBM Research paper by Murray Campbell, A. Joseph Hoane Jr. and Feng-hsiung Hsu. That documentation makes it possible to study the system without treating the event as magic.

Kasparov’s response also changed over time. He later wrote and spoke about cooperation between people and machines, including “advanced chess,” in which human players use computer assistance. The idea shifted the question from whether a machine could defeat a person to how combined strengths might solve problems.

What Deep Blue Did—and Did Not—Prove

Chess had long served as a benchmark for computing because its rules are clear but its possible positions are immense. Defeating the world champion showed that specialized hardware and software could outperform the strongest human in that domain. It did not prove that the computer thought, felt pressure or understood victory as a person would.

The match nevertheless influenced public expectations. Later milestones—IBM Watson on Jeopardy!, AlphaGo’s victory over Lee Sedol and modern generative systems—were often described through the same human-versus-machine frame. Each used different methods and addressed different tasks.

May 11, 1997 remains valuable because the claim can be stated precisely. Deep Blue won a regulated six-game chess match, 3½–2½, after losing to Kasparov the previous year. Its triumph combined fast custom hardware, expert-built software and human preparation. For more technology and cultural milestones around it, see the full 1997 US calendar.