The history of Go: 2,500 years, unchanged rules

Origins in China: the oldest game still played

Go is the oldest board game that is still played with essentially unchanged rules. The first reliable written reference is in the Zuo Zhuan, a Chinese chronicle describing events of 548 BCE, where a politician's indecision is compared to a Go player who holds a stone without knowing where to put it. The metaphor only works if the game was already familiar.

Legend goes further back: the Emperor Yao is said to have invented Go to teach his son discipline. That story first appears many centuries after the fact and is best treated as a founding myth rather than history. What the archaeology and texts support is a game established in China by the middle of the first millennium BCE, played on a board that grew from 17×17 to the 19×19 we use today.

Confucius mentions it. The Han dynasty produced essays on it. By the Tang dynasty it was one of the four arts of the Chinese scholar — alongside the qin (zither), calligraphy and painting — which is the single most important fact about Go's early history: it was not a pastime, it was an accomplishment expected of an educated person.

Korea and Japan

Go travelled east along the same routes as Buddhism and the Chinese writing system. It reached the Korean peninsula early — Korean baduk later developed its own historical variant, sunjang baduk, with a fixed opening arrangement of stones — and arrived in Japan around the seventh century CE.

Japan is where Go was professionalised. In the early seventeenth century the Tokugawa shogunate began supporting Go houses — Honinbo, Inoue, Yasui and Hayashi — with stipends, and instituted the oshirogo or Castle Games, annual matches played in the presence of the shogun. The strongest player held the title of Meijin. For 250 years Japan had something no other game anywhere had: a state-funded professional structure with a documented competitive record.

Honinbo Shusaku and the golden age

Honinbo Shusaku (1829–1862) is the most famous player in the game's history. He is remembered for two things: an unbeaten run in the Castle Games, and the "ear-reddening move" of 1846 — a move so good that a watching doctor, who did not play Go, said he knew Shusaku had won because his opponent's ears turned red.

The Shusaku opening — a specific sequence of Black's first moves — dominated play for a century and was still standard when the twentieth century began. It only lost its status when komi was introduced, which changed the value of Black's first move and made the old slow openings too generous to White.

The house system collapsed with the Meiji Restoration when the stipends ended. The modern structure that replaced it — the Nihon Ki-in, founded in 1924 — invented the newspaper-backed title match, which is still how professional Go is funded in Japan.

Go Seigen, Korea and the Chinese return

In 1933 Go Seigen — a Chinese prodigy who moved to Japan — and Kitani Minoru published a new approach called shin fuseki, "new opening", which threw out centuries of convention about where to begin. Go Seigen then spent two decades defeating every top Japanese player in a series of jubango (ten-game matches), and is usually named the strongest player of the twentieth century.

From the late 1980s the centre of gravity moved to Korea. Cho Hunhyun's victory in the first Ing Cup in 1989 began a period of Korean dominance carried on by Lee Chang-ho, whose endgame precision reset the standard for the whole game. From the 2000s China returned to the top, and today the strongest human players are overwhelmingly Chinese and Korean.

Why computers took twenty extra years — and then AlphaGo

Chess fell to computers in 1997. Go held out until 2016, and the reason is arithmetic. A 19×19 board has roughly 10¹⁷⁰ legal positions — more than the number of atoms in the observable universe, squared — and, worse, no simple way to tell whether a position is good. In chess you can count material. In Go, a stone's value depends on a whole-board judgement no simple formula captures.

Two breakthroughs cracked it. Monte Carlo tree search (from around 2006) replaced evaluation with random playouts: instead of judging a position, play it out thousands of times and see who tends to win. Then deep neural networks gave programs the intuition MCTS lacked.

In March 2016, DeepMind's AlphaGo beat Lee Sedol, one of the strongest players of his generation, 4–1 in Seoul. AlphaGo's move 37 in game two — a shoulder hit on the fifth line that no professional would have considered — is now the standard example of a machine producing a genuinely new idea in a 2,500-year-old game. Lee's move 78 in game four, which won him the only human game of the match, has its own name: the "divine move".

A year later AlphaGo Zero learned the game from nothing but the rules and self-play, and surpassed every previous version within days.

How Go reached the West

Go arrived in Europe late and in a roundabout way. The first substantial Western account was written by Oskar Korschelt, a German chemist working in Japan in the 1880s, whose treatise introduced the game to German readers and, indirectly, to the rest of Europe. Interest grew slowly through chess circles, where the game was received as an intriguing curiosity rather than a rival.

The most effective early advocate in the English-speaking world was Edward Lasker, an engineer and chess master who learned Go in Berlin, emigrated to the United States and helped found the American Go Association in 1935. His book Go and Go-Moku was many Americans' first contact with the game.

Two later waves did far more than any book. The first was the European Go Congress, held annually from 1957, which gave the continent a real playing community. The second was the internet: Go servers from the mid-1990s let a beginner in a town with no club play ranked games against opponents anywhere, which is still how most Western players learn.

Go today

Go is played by tens of millions of people, concentrated in China, Japan, Korea and Taiwan, with active communities across Europe and the Americas. Professional players are salaried and compete for title matches that carry names — Honinbo, Meijin, Kisei in Japan; Kuksu and Myeongin in Korea; the Ing Cup and Samsung Cup internationally.

The game itself has not changed. A Han-dynasty player set down at a modern board would need to be told about komi and would find the openings strange, but would be playing the same game within a move or two. Almost nothing else humans do has that continuity.

Common questions