Kasparov vs. Deep Blue, 1997

Kasparov had faced Deep Blue's predecessor in February 1996 in Philadelphia and won the match 4–2. But the computer won Game 1 of that match, an unprecedented result, and Kasparov later admitted it shook him psychologically. The 1996 Deep Blue calculated roughly 100 million positions per second. The 1997 version, rebuilt and reprogrammed, was significantly stronger, though IBM refused to disclose exactly how much faster. See the full Game 1, 1996.

This secrecy would become a central flashpoint. Kasparov requested access to Deep Blue's previous games for preparation (standard practice when facing a human opponent), but IBM provided only a limited set. The asymmetry, the machine's team could study every Kasparov game ever played, while Kasparov couldn't study the machine, grated on him throughout the match.

Game 1 went to Kasparov. He played the King's Indian Attack, steered into a complex middlegame, and won convincingly. It seemed like a repeat of 1996. Then came Game 2, the game that changed everything.

In a Ruy Lopez, Deep Blue played a piece sacrifice on move 36 that Kasparov couldn't understand. The move, 36.axb5, was objectively strong, but its purpose was opaque. Kasparov, visibly disturbed, played passively and eventually resigned in a position that later analysis showed was actually drawn. Some observers believe Kasparov mistakenly thought the computer was seeing something deeper than it was; the reality was probably a combination of genuine computer strength and Kasparov's own psychological fragility. See the full Game 2, 1997.

Game 2 broke Kasparov. He later said that the computer's play showed "a quality of deep intelligence" that he couldn't attribute to mere calculation. Whether that quality was real or imagined became the most debated question in chess-AI history.

After losing the match (Deep Blue won Games 2 and 6, Kasparov won Games 1 and 3, with two draws), Kasparov demanded a rematch. See the decisive Game 6. IBM refused and retired Deep Blue. They never released the machine's full logs. This decision fed conspiracy theories, that human intervention had assisted the computer during the match, that IBM's real goal was a stock-price spike (IBM shares did rise significantly), that the match was rigged.

None of these theories have been substantiated. The most likely explanation is prosaic: IBM achieved its publicity goal, a rematch carried only risk with no reward, and a corporate decision was made to move on. But the refusal to allow transparency left a stain that persists to this day.

Kasparov vs. Deep Blue did not end chess. It did not make human competition meaningless. What it did was accelerate a transformation that was already underway: the integration of computer analysis into every aspect of chess preparation, study, and improvement.

Today, every serious player uses an engine. The game analysis you run after a game on ChessBhumi is powered by Stockfish, an engine that would crush Deep Blue in every game without effort. The barrier that Deep Blue crossed in 1997 has been so thoroughly surpassed that the question of "can a computer beat a human?" is no longer interesting. The interesting question now is: how do computers make humans better?

Use engine analysis to review your own games. The engine will show you the moves you missed, the tactical combinations you didn't see, and the strategic ideas you didn't consider. That's the real legacy of Deep Blue: not the defeat of a champion, but the democratisation of top-level chess insight.