AI Perfected Chess. Humans Made It Unpredictable Again
Artificial intelligence drove chess toward perfect play, leading to more draws at top tournaments. Now grandmasters are winning by making less optimal moves.
Illustration: Mathieu Labrecque for Bloomberg
At the biannual 2018 World Chess Championship, Magnus Carlsen — commonly regarded as the best chess player to ever live — defended his title against challenger Fabiano Caruana in a best-of-12 format. Classical chess allows players long stretches of time in which to make their moves, and the two logged more than 50 hours of play across 12 games. To the shock of the chess world, every single game resulted in a draw, a first in the history of the championship, which dates back to 1886. (Carlsen went on to win after three tiebreakers.)
This seemed to confirm a growing suspicion: Chess was dead — and draws had killed it. The “draw death” of chess was not a new fear. In 1925, then-World Champion José Raúl Capablanca became so worried about the increasing sophistication of the game’s top players — whom he believed were not far away from drawing games at will — that he proposed a new set of rules to save chess. But it never quite happened: Over the ensuing decades, the draw rate between masters playing classical chess hovered around 50%.