Deep Blue versus Garry Kasparov

Deep Blue

Company type	Computer chess system
Industry	Computer chess, Artificial intelligence
Founder	IBM
Headquarters	United States
Area served	Worldwide (chess competitions)
Products	Chess-playing computer system
Owner	IBM

Garry Kasparov, who was the world chess champion at the time, engaged in two six-game encounters against IBM's Deep Blue. Kasparov emerged victorious in the initial 1996 match held in Philadelphia with a score of 4–2. However, in the 1997 New York rematch, Deep Blue triumphed 3½–2½, marking the first occasion a reigning world champion was defeated by a computer in official tournament conditions. This historic event was documented in the film Game Over: Kasparov and the Machine.

Impact and symbolic significance

The matches attracted widespread media attention, significantly raising Deep Blue’s public profile. Despite reports that IBM dismantled Deep Blue following the contests, the system continued to function for several years afterward.

IBM Research provided prize funds for both competitions, with all earnings awarded to IBM. The initial match offered $400,000 to the victor and $100,000 to the runner-up; the subsequent match increased these amounts to $700,000 and $400,000 respectively. Additionally, Carnegie Mellon University contributed $100,000 to Deep Blue’s team through a prize founded by computer science professor Edward Fredkin, awarded to the first program capable of defeating a reigning world chess champion.

Deep Blue’s success came to symbolize a significant step forward in artificial intelligence achieving human-like cognitive abilities. Later evaluations tempered this interpretation by noting that Kasparov’s defeat partly resulted from unusual mistakes on his part and raised doubts about whether chess alone serves as an adequate measure of intelligence due to its vulnerability to brute-force computational methods.

In a 2016 podcast, Kasparov shared how his perspective had evolved after conducting research and writing on the subject. He expressed greater admiration for Deep Blue’s creators and acknowledged that contemporary chess engines now vastly outperform Deep Blue’s capabilities.

Following Deep Blue’s breakthrough, the ancient game of Go—which features exponentially more possible moves and demands intuitive judgment—became the new frontier for AI challenges. In 1997, top Go programs still lagged behind amateur players. Yet Google DeepMind’s AlphaGo demonstrated a major advancement by defeating European champion Fan Hui in 2015 and then world-class player Lee Sedol in 2016. Unlike Deep Blue’s brute-force search through millions of positions, AlphaGo employed neural networks combined with reinforcement learning, ushering in a novel era of AI strategy formulation.

Summary

The 1996 match

Game #	White	Black	Result
1	Deep Blue	Kasparov	1–0
2	Kasparov	Deep Blue	1–0
3	Deep Blue	Kasparov	½–½
4	Kasparov	Deep Blue	½–½
5	Deep Blue	Kasparov	0–1
6	Kasparov	Deep Blue	1–0
Result: Kasparov–Deep Blue: 4–2

The 1997 rematch

Game #	White	Black	Result
1	Kasparov	Deep Blue	1–0
2	Deep Blue	Kasparov	1–0
3	Kasparov	Deep Blue	½–½
4	Deep Blue	Kasparov	½–½
5	Kasparov	Deep Blue	½–½
6	Deep Blue	Kasparov	1–0
Result: Deep Blue–Kasparov: 3½–2½

Game 1

. The opening encounter featured the Sicilian Defence, Alapin Variation. This match marked the first instance where a computer defeated a reigning world chess champion under official tournament rules and classical time controls.

Game 2

. The second game shifted into an open variation of the Catalan Opening. Kasparov employed a proactive approach, hindering Deep Blue’s attempts to develop its pieces. After 73 moves, Deep Blue conceded defeat; although opposite-colored bishops often lead to drawn outcomes, Kasparov’s trio of connected passed pawns proved superior to Deep Blue’s solitary passed pawn.

Due to a setup error, Deep Blue played this game without access to its opening book but was still able to replicate established opening theory through move 13 by utilizing its extensive database of 700,000 grandmaster games.

Game 3

. Kasparov initiated with the Sicilian Defence, countered by Deep Blue’s Alapin Variation. The contest concluded in a draw after 39 moves.

Game 4

. The fourth encounter ended in a draw despite Kasparov’s mid-game offer being declined by Deep Blue’s team. The opening followed a variation of the Queen's Gambit Declined.

Game 5

. In game five, Deep Blue rejected Kasparov’s draw proposal after move 23 but ultimately suffered defeat—the sole victory for Black throughout the match. The opening employed was the Scotch Four Knights Game, combining aspects of both the Scotch Game and the Four Knights Game.

Game 6

. The sixth game echoed the fourth by adopting the Queen's Gambit Declined. Kasparov implemented strategies aimed at countering computer play, concentrating on long-term positional improvements while Deep Blue struggled to make progress. By the conclusion, Deep Blue’s pieces were restricted to its queenside corner with minimal mobility, enabling Kasparov to gain a decisive upper hand. Kasparov was set to exchange queens and advance a passed pawn toward promotion, securing a winning advantage.

Game 1

. The rematch began with the Réti Opening, which evolved into the King's Indian Attack. Kasparov emerged victorious after 45 moves.

Deep Blue’s move on turn 44 was reportedly caused by a software glitch that activated a fail-safe mechanism when no optimal move could be identified.

Game 2

The second encounter commenced with the Ruy Lopez Smyslov Variation. Kasparov ultimately conceded defeat, although post-game analysis suggests he might have secured a draw. Following the match, Kasparov accused IBM of cheating, asserting that human assistance influenced a crucial move—an allegation explored in the documentary Game Over: Kasparov and the Machine.

Deep Blue’s initial nine moves adhered to its opening repertoire; subsequent plays were drawn from an expanded opening book up to move 19...Nh4, after which the game proceeded beyond known theory.

Analysts have pointed out that after move 45...Qe3 and the ensuing sequence, Black had an opportunity to force a draw through perpetual check. Critics observed that Deep Blue’s king maneuver at move 44 was less than ideal, enabling this option. Chess commentator Mig Greengard and four-time US champion Yasser Seirawan identified this as a significant oversight by Deep Blue, implying Kasparov could have avoided defeat by compelling a draw.

Kasparov’s claim of cheating focused on moves such as 36.axb5! axb5 and 37.Be4!, which diverged from the typical material-focused approach expected from computer play. Instead of pursuing material gains with moves like 36.Qb6 or 37.Qb6, Deep Blue opted for more nuanced positional strategies, surprising Kasparov and others who considered such sophistication beyond machine capability at the time, thereby intensifying suspicions of human involvement during the match.

Game 3

Kasparov initiated the third game with the unconventional 1.d3, known as the Mieses Opening, which transitioned into a variation of the English Opening. His plan was to steer Deep Blue out of its prepared opening lines, anticipating weaker performance without its standard repertoire. Despite this approach, the game concluded in a draw.

Game 4

The fourth game began with the Caro–Kann Defence (1.e4 c6) but soon transposed into the Pirc Defense. Facing time constraints in the latter stages, Kasparov made less precise moves that likely deprived him of a potential victory. The game ended in a draw.

Game 5

Starting similarly to Game 1 with the Réti Opening evolving into the King's Indian Attack, Game 5 featured Deep Blue demonstrating strong endgame technique to hold a draw despite Kasparov’s winning prospects. It was later disclosed that Kasparov had a winning continuation beginning with 44.Rg7+.

If White promotes with 50.g8=Q, Black can enforce a draw via threefold repetition starting with 50...Rd1+.

Game 6

The final and decisive sixth game was the briefest in the series. With the overall score tied at 2½–2½, Kasparov once again employed the Caro–Kann Defence. Deep Blue’s knight sacrifice shattered Kasparov’s defenses, compelling him to resign in fewer than twenty moves. Kasparov acknowledged that he selected a risky opening intended to destabilize Deep Blue but underestimated the engine’s tactical precision.

See also

• AlphaGo versus Lee Sedol
• Arimaa – Inspired by Kasparov's defeat to Deep Blue, this game was created to challenge computers while playable on a standard chess set.
• List of chess games
• Rematch – A 2024 television miniseries dramatizing the 1997 rematch

Primary sources

• Pandolfini, Bruce (1997). Kasparov and Deep Blue: The Landmark Chess Contest Between Human and Machine. Fireside Chess Library. ISBN 9780684848525.
• Garry Kasparov versus Deep Blue, 12 games, Chessgames.com

Additional reading

• IBM Research. Official Deep Blue match website.
• Gardner, Martin (2001). A Gardner's Workout: Exercising the Mind and Enriching the Spirit. A K Peters. ISBN 978-1-56881-120-8. pp. 91–95.
• Goodman, David; Keene, Raymond (1997). Man Against Machine: Kasparov Versus Deep Blue. H3. ISBN 978-1888281064.
• Hsu, Feng-hsiung (2002). Behind Deep Blue: The Creation of the Computer That Defeated the World Chess Champion. Princeton University Press. ISBN 0-691-09065-3.
• Khodarkovsky, Michael; Shamkovich, Leonid (1997). A New Era: World Championship Chess in the Deep Blue Epoch. Ballantine. ISBN 978-0345408907.
• King, Daniel (1997). Kasparov versus Deeper Blue: The Definitive Man versus Machine Challenge. Batsford. ISBN 0-7134-8322-9.
• Newborn, Monty (1996). Kasparov versus Deep Blue: The Maturation of Computer Chess. Springer. ISBN 0-387-94820-1.