AI Horizon: Intelligence and Reason in Computers
Introduction Home Essays Resources Contact Info
  Basic CS
General AI
Chess AI

  Source Code
AI Tips and Tricks

  Code Journal
Site Updates


Introduction to the Game of Go and the Rules
(Up to Go AI)

A Brief Background

Go (also Igo) is the Japanese name for an ancient board game that originated in China between 2000 to 3000 years ago. The name for the game in Chinese is weiqi (pronounced way-chee). This means, roughly, "surrounding game." In Korean, the name is Baduk (pronounced bah-dook, not bah-duhk). In Taiwan, the game is referred to as "Goe."

The game spread from China to Korea, and then it most likely spread to Japan from Korea. Stories talk about a Korean buddhist monk who sailed the ocean to bring the game to the Japanese royal courts. The modern rules and strategy were developed in Japan. Today, the game is played by people all over the world. Some estimate that more than 100 million people play this game.

The Basic Rules

NOTE: These are the main rules used in modern Japan and Korea.

The main goal in Go is to surround more unoccupied areas of the game board ("territory") than the opponent. Unlike Chess, the opponent is not eradicated or destroyed. Rather, the two players seek to coexist on the board, leaving one player with more.

A Standard 19 x 19 size boardThe game is usually played on a 19 x 19 grid, which is the standard tournament size, but other sized game boards are used as well. The most popular alternative sizes are 9 x 9 and 13 x 13. Boards smaller than 9 x 9 are usually too small and cramped, and boards larger than 19 x 19 are usually too unwieldy. The dots are simply visual reference points, and do not affect play.

There are two players, one playing White and the other playing Black. Black moves first, except in handicap games, and the players play on alternating turns. Each turn, a player places one piece (a "stone") onto an intersection (not the square) of the board. Edges and corners are also playable intersections. The intersections are called "points." Once placed, the stones are never moved again, except when captured, which will be explained shortly.

A picture showing libertiesA played stone has a certain number of liberties. The liberties are the number of lines extending from a stone. In diagram, which shows a partial view of a go board, stone A has four liberties, stone B has three, and stone C has two. D would be one of the liberties of A.

A group of adjacent friendly stones (stones of the same color) share liberties. That group is called a string. The stones, however, must be directly adjacent in order to form a linked group; diagonals do not count. So, if a black stone were to be at D, then A and D would form a string with six liberties total. The two stones are now an inseparable unit, a string.

The triangled stones are capturedLiberties are reduced by adjacent enemy stones. If a player plays a stone such that an enemy group has no more liberties (the enemy is "surrounded"), that player is said to have captured the enemy group, and the enemy group is removed from the board and given to the player to hold as prisoner. Each prisoner piece counts as one point towards the score of the player that holds it. In the picture, White has captured the four triangled Black stones, which are removed as White's prisoners. Stone A is not captured, and it is separate from the other Black group.

A and B are illegal moves for WhiteThere are only a few illegal moves in go, but one of them is this: most directly suicidal moves are illegal. In the picture to the right, both A or B are illegal moves for White. They are perfectly legal for Black.

A finished 9 x 9 gameThe game essentially ends when both players agree that the score will not change any more. This is generally when the borders have all been solidified. The score for each player is counted as follows: count the number of empty intersections that the player controls. To that number, add the number of prisoners that player holds. This is the raw score of the player. Traditionally -- since Black has the advantage of the first move -- White is given a certain number of points, usually ranging between 5.5 to 7.5, depending on the tournament. The most common komi is 6.5 points. This is called komi in Japanese. (The half point is included to prevent ties.) Assuming that there are no prisoners, in the finished game shown to the left, Black has 32 points, and White has 28.5 (assuming a 6.5 point komi). So, Black wins by 3.5 points.

The Ko Rule

The KoThe most complicated rule is the rule for ko, which is a Japanese word meaning something like "everlasting." Consider the following position to the right. Let's say that it is Black's turn. If Black plays at A (this is legal because Black is capturing White's stone at B, which means that his stone he played still has one liberty), then White could recapture at B (Black captured the White stone B, so the point is now empty). Then, Black could recapture at A, and White could...

This could go on forever. So, there is a rule to prevent such everlasting repitition. This formation (and its basic variations) are known as the ko, so the rule is known as the ko rule.

If Black or White captures in a ko, then the opponent can't just recapture. Instead, the opponent has to play elsewhere first, and only later can the opponent return to capture the ko (assuming that the first player doesn't connect the ko).. Often, this "in between move" is a threat (to capture, or do something damaging) which tries to cause the first player to answer the threat instead of connecting the ko. In this case, this threat is known as a ko threat.

Life and Death

This last topic is not really a rule, but a crucial consequence of the rules. Because suicide is illegal, the following Black group is uncapturable, no matter what White tries to do.

Both of the empty spaces inside the Black group are illegal moves for White, and since White can only play one move at a time, White can never capture Black (unless Black feels stupid, a bit generous, or both and fills in one of the spaces). Each of the two completely surrounded spaces is called an eye, and any group that has at least two eyes is alive. The opposite of life, of course, is death. Any group, such as the following Black group, that can't make two eyes is dead.

Often, players will simply allow dead groups to stay on the board until the end of the game (without actually capturing them), with the understanding that they are dead and that the capturing details don't actually need to be worked out. This is often true of small scattered stones in enemy territory. After the game is finished and counting is to begin, the dead groups are removed and simply counted as prisoners.

This life and death aspect adds whole new strategic and tactical dimensions, and computers are still not able to solve even this small - but important - portion of the game as well as top human players.

With some practice, the rules of the game become clearer and easier to understand instinctively. There are many basic computer programs available that allow you to practice playing go. The rules usually only take a few hours to master completely, but the strategy is complex enough that several millenia of study has still not solved the game. The visual nature of the game, as well as the vastness of its possibilities, makes it difficult for any computer to play it to any high degree of ability.

This game is a double blessing in that humans will reign as champions over computers for decades, perhaps centuries to come, but also in that this game provides an active and controlled environment in which to test exciting new techniques for Artificial Intelligence.

All content is written and published by the people at or affiliated with AI Horizon <>.
Send any comments and suggestions to

Please report any errors to