Go Game Tree Size Exceeds a Googleplex: Key Takeaways

The game of Go, an ancient East Asian game popular in China, Japan, and Korea, has been played for approximately 2,500 years with remarkably consistent rules. Beyond its historical significance, Go presents fascinating mathematical properties, particularly concerning the vast numbers associated with its gameplay.

The Googleplex and Real-World Numbers

The discussion centers on the concept of a Googleplex, defined as 10 to the power of a Google (10^100). This number is so immense that it holds a place in the Guinness Book of Records as the largest finite number with a widely accepted name, appearing in the Oxford English Dictionary. While such colossal numbers are typically encountered in pure mathematics, Go offers a rare instance where numbers of this magnitude can be found in a real-world context.

Basic Rules of Go

Go is played on a 19x19 board, significantly larger than chess's 8x8 board, which contributes to the vastly greater number of possibilities in Go. Players place "stones" on the intersection points of the grid, not within the squares.

Capturing Stones

A fundamental rule involves the capture of stone groups. A group consists of stones connected horizontally or vertically. If a group of stones is completely surrounded by an opponent's stones, they are captured and removed from the board. For example, if a black group is surrounded by white stones, and white plays the final stone to complete the encirclement, the black stones are captured. The spaces they occupied then become available again.

The Super Ko Rule

A more subtle but mathematically crucial rule is "super Ko." This rule prevents a game from continuing indefinitely by prohibiting any move that would result in the board returning to an identical previous state. While not frequently encountered in practical games, super Ko is vital for mathematical analysis as it guarantees that every game will eventually end, ensuring a finite number of possible games.

Winning the Game

Winning in Go is about controlling territory. Players can pass their turn. If both players pass consecutively, the game ends. The winner is determined by who controls more territory, which includes both the stones on the board and the empty intersection points enclosed by a player's stones.

Mathematical Analysis of Go

Number of Board Positions

The number of possible board positions in Go was first investigated in the 11th century by the scholar Shen Kuo. Each of the 361 (19x19) intersection points can be in one of three states: black, white, or unoccupied. This leads to 3^361 possible configurations, which is approximately 2 x 10^172.

However, this initial count includes many "illegal" positions. An illegal position occurs when a group of stones is depicted as being surrounded and still on the board, even though they would have been captured and removed. Research by John Trump and his collaborators in 2016 determined that only about 1% of these configurations are actually valid. Even with this reduction, the number of valid board positions remains enormous, around 2 x 10^170.

Number of Possible Games

To reach even larger numbers, one must consider the number of different possible games, which involves stringing together valid positions according to the rules. This is often referred to as the "game tree." The super Ko rule is essential here, as it ensures a finite number of possible games by preventing infinite loops.

Analysis on a 2x2 Board

To illustrate the complexity, consider a simplified 2x2 Go board. On this tiny board, there are 3^4 = 81 possible configurations. Of these, 57 are legitimate. Despite the small size, the number of possible games is surprisingly large. John Trump's work shows that there are over 386 billion (386,356,999,593) possible games on a 2x2 board. This high number is due to the ability of stones to be captured and re-played, leading to long game sequences.

An example of a long game on a 2x2 board, lasting 48 moves (including passes), demonstrates how cycles of capture and recapture, slightly altered each time to avoid the super Ko rule, can extend game length significantly.

Number of Games on a 19x19 Board

Extrapolating from the 2x2 board, the number of possible games on a full 19x19 board is astronomically larger. As of 2016, John Trump and Matthew Volart proved that there are at least 10^(10^108) distinct games on a 19x19 board. This number surpasses a Googleplex, making Go one of the few real-world phenomena where such immense numbers are relevant.

It's important to note that most of these theoretically possible games are unrealistic, involving an impractical number of moves that no human player would ever complete.

Realistic Games and Perfect Play

Estimating the number of "realistic" games involves making assumptions about average game length and the number of available moves per turn. If an average Go game has 200 moves and approximately 250 moves are available at each turn (compared to about 30 in chess), the number of realistically possible games is roughly 250^200, which is about 10^500. While significantly smaller than a Googleplex, this is still a colossal number, dwarfing quantities like the number of atoms in the universe.

Another question is how many games are possible assuming "perfect play." While computers can now beat human Go champions, a perfect strategy for Go (or even chess) remains unknown and is likely an impossibly distant prospect due to the immense complexity and vast number of possible game states.

The simplicity of Go's rules belies an immense underlying complexity, making it a rich subject for mathematical exploration and a rare example of a real-world system that generates numbers on the scale of a Googleplex.