Rules of go - second tutorial
This page presents various details of the rules and their history for those who are interested in such things.
- If just want to know how to play, read Rules of Go - introductory. This page aims to answer questions that may occur to intellectually curious players; thus it does not assume a very sophisticated understanding of the game, but does assume the reader is interested in rules trivia as well as theoretical and in practice unusable concepts. This page also aims to present formal concepts that are either characteristic of how experienced players think about the game, or can be used to make the rules of Go perfectly consistent and to clarify why different rule sets exist.
|Table of contents|
A game can end in a tie, with both players having the same score. This is called jigo in Japanese.
White has captured three Black stones (at the points marked ), and has five points of territory.
Black has eight points of territory.
On a 19x19 board, the White player normally gets 6.5 points as compensation for moving second. These points are added to white's score at the end, as though they were additional points of territory. This compensation is called komi in Japanese.
Komi usually includes a half-point to prevent ties.
The exact number of points of komi varies from rule-set to rule-set. Komi may also vary by board-size; many players feel that White needs some compensation on a 9x9 board, but that 6.5 points is too much.
Go is a deep game. The more you play, the more depth there is to explore. This means that, compared to a card game or a casual party game, experienced Go players will consistently beat less experienced players in an even game. In fact, a beginner who has four or five games under his belt will consistently beat a new player, but that same beginner will consistently lose to a slightly more advanced beginner who has played a few dozen games: and so on up the line.
You might think this would ruin the fun of the game, or at least make it difficult to find an opponent of the same level. But in fact, Go has a remarkably accurate handicap system that allows players of different ability to play with a reasonably even chance of winning.
Two players who are close in strength will play an even game, that is, a game where Black plays first and White receives komi. If one player can beat another approximately 2/3 of the time in even games, the stronger player gives the weaker a handicap of one stone; that is, the weaker player always takes Black and plays first, but the stronger player does not receive komi. If this handicap is not enough, then Black can put down two, three, or as many as nine stones on the board. Traditionally the handicap stones go on the star points (after which White plays the first move), but White can also permit Black to place his handicap stones in some other formation.
It is now normal for White to receive 0.5 points of komi even if White does not need compensation to make the game fair, to formalize the convention that a tie counts as a victory for white.
It is worth noting two interesting features of the Go handicap system. First, even with a very high handicap the local situation will resemble a position that could arise naturally in an even game. The game flows in a natural and familiar manner, keeping the game fun even with a high handicap. Second, the handicap system is transitive; that is, if Alex gives Bob five stones, and Bob gives Carla three stones, eight stones will probably be an appropriate handicap for a game between Alex and Carla. This allows all Go players to be integrated into a common ranking system in which differences in game experience are measured in handicap stones.
In the introductory rules, the goal of the game was stated concisely: Go is about controlling as much of the board as possible. But what is meant by “control” could be vague. You could say that you control any space where you could play a stone; or that you control any space where your opponent cannot play a stone; or perhaps something different from either of these two. It turns out that precisely how you state the goal of the game affects the score.
Originally, the goal of Go was to place as many of your stones as possible on the board. This required filling the board (almost) completely with stones after the boundaries of the territories had been settled. The monotony of this filling-in led players to invent mathematical shortcuts to calculate the score earlier in the game. However, some of these shortcuts came to seem unnatural once players were no longer used to thinking of filling the board with stones as the goal of the game, and were dropped.
In the original form of Go, called Stone scoring, only stones on the board count towards victory. Thus in effect, a player does not control the eyes of his group (the two liberties each group will have at the end of the game). Compared to modern Go, this concept of control levies a tax of two points on each group. (Or, compared to ancient Go, all modern forms of Go give an “eye bonus”: two extra points for the eyes of each group.) If Black and White have different numbers of groups on the board at the end of the game, a rule which assesses a group tax will give a different score than a rule which awards an eye bonus.
Dame are neutral points between territories, and dame parity refers to the question whether the number of dame is odd or even. If the dame parity is odd, this can cause different scoring systems to give different results.
Dame are not counted as points under most scoring systems, but some scoring systems count the stones on the board. When all other possible moves have been made and only dame remain, they will be played alternately by both players. If there are an even number of dame, each will get exactly half of them. If, however, there are an odd number of dame, the player who starts filling them will get one more point from the dame than their opponent; in these cases (about half of all games), counting stones on the board rather than territory will affect the overall score.
In modern Go, these two approaches are usually referred to as “territory scoring” and “area scoring”. However, note that rule sets using area scoring generally have higher komi (currently, 7.5 rather than 6.5), to compensate for dame parity.
PJT: This seems a little odd, given that the average difference is ˝.
In ancient Go, the game ended with the board completely filled with stones, so that there was nothing left to argue about. Because modern Go does not require filling in the board, it does require players to agree which stones are dead, i.e. can be captured, and to remove them from the board before tallying the score. However, if players do not agree on the status of dead stones – a problem that frequently affects new players – a mechanism is needed to resolve such disputes.
Players unfamiliar with it are often worried that the score could change during the dispute resolution phase. This cannot happen. All modern Go rules are modifications of the ancient stone scoring rules, and are designed to give similar results.
The concept of control used in area scoring does not distinguish between stones and territory, so (as in stone scoring) one can play on to the bitter end, capturing all the dead stones and removing them from the board, without affecting the score. The concept of control used in territory scoring punishes players for making extra moves inside their own territory, so here status disputes at the end of the game are resolved by a special mechanism specified by the rule set:
- They may play a so-called encore: an extra period of play under slightly different rules about repetition producing a new position to be scored.
- They may use hypothetical play, i.e. discussion of the possibilities, to establish the status of all stones in the existing position.
- In this case, weaker players may actually play out sequences on the board, in which case care should be taken to ensure the board is returned to its state when normal play stopped.
- See also the discussion under encore for the difference
Patrick Traill (2020-08-15) The following statement seems too strong (the players may make mistakes) and no longer fits the description of three possibilities!
If dame parity affects the score, then all of the stones actually present on the board need to be tallied to determine how many points a player has. Beginners often do this manually on small boards, but there is a technique for calculating the score quickly on a big board. If dame parity does not affect the score, then it can be tallied with captured stones and territory only; once the captured stones have been used to fill in the opponent's territory, the remaining territories are small and relatively easy to count.
In practice, two main concepts of control are territory scoring and area scoring. Territory scoring is used by Japanese and Korean professionals, many Western amateurs, and most internet servers. Area scoring is used by Chinese professionals, the official tournaments of most Western go associations, and a few Western amateurs. Which rules are easiest for beginners to learn, encourage the best habits of play, and lead to the best games is a source of quiet but stubborn disagreements.
Life and death is not precisely a rules concept, but in practice it is very important not only to Go strategy, but also to when experienced players pass, ending the game.
As a strategic concept, life and death is important because, as more stones are placed on the board and liberties disappear, some groups will inevitably lose all their liberties and be captured, while other groups can never lose all their liberties and thus will stay on the board at the end of the game. Since a Go player does not want his opponent to capture his stones, he must make them uncapturable, and this requires that they end the game with at least two liberties. If these two liberties are adjacent to one another, the opponent can normally reduce the group to one liberty by playing on one of the two; so in general these last two liberties must also be separated by stones. The separated liberties are called eyes; having two eyes, or being able to make two eyes is sufficient for a group to live. However, a group whose last two liberties are adjacent might also be uncapturable, if the same two points are also the last two liberties of an enemy group. In that case, if either player fills in the first of the two shared liberties, his opponent will fill in the second and capture the group. This is called mutual life or seki.
At the end of the game, each player allows his opponent to remove the dead stones inside his own territory and add them to his territory. If a player believed those dead stones had a chance to live, he would have saved them before passing; if he knows they did not have a chance to live, he does not bore his opponent by forcing him to capture them.
Without the concept of life and death, or without a good understanding of when stones are alive or dead, it is still perfectly possible to play a game of Go, but it might make the end of the game laborious.
See: Life and Death
The ko rule (the rule forbidding repeated board positions) was instituted to deal with the normal mid-game ko situation. However, this rule can be ambiguous in unusual situations, and many revisions have been proposed.
Many of these situations are too complex for beginners to understand. Consult the page on ko for more consideration.
According to the traditional rules of Go, one may play a stone in a position where it has no liberties (or is a part of a group that has no liberties) if and only if the move captures enemy stones which, when removed from the board, leave behind empty spaces that provide the stone with liberties.
A common confusion among beginners has it that this rule "prohibits suicide". Some even hold that the traditional rule is an extraneous and inelegant wart on the core rules of the game. However, the truth is more complicated. The traditional rules permit playing stones without liberties that make captures, but make no provision for what would happen if a stone were without liberties at the end of it's own players turn. (This was in number 7 capturing of the last tutorial "If there are no empty points next to a stone or a string of stones (the stone has no liberties), the stones are immediately taken off the board." so it seems it is already settled. The pieces would be removed –Note from Keith M. McCabe?)
The easiest way to "permit" suicide is by adding a capture sequence rule, to the effect that any stone can be placed without liberties; then the opponent's groups are checked to see if any of them have no liberties, and can be captured by the moving player; next, the moving player's stones are checked, and can be captured by the opponent. (Note that this has the slightly undesirable consequence that a player will always have a legal move other than a pass, so the game need not end.) This capture rule is officially part of the rules of the New Zealand Go Association. Of the variant capture rules, this one is the most similar to traditional Go; it sometimes creates additional ko threats and otherwise does not affect game play.
But one could equally well permit any stone to be placed without liberties; check the moving player's stones for capture first; and then check the opponent's stones for capture. This rule would fundamentally change life and death.
Or, one could permit any stone to be placed without liberties, and only remove stones that lack liberties at the end of the opponent's turn. This would allow a stone or group of stones to stay on the board without liberties through out an opponent's entire turn, at the end of which they would be captured. This would change the life-and-death status of some groups.
One could also permit the placement of a stone without liberties, and remove all stones lacking any liberties at the same time. This would, for example, change the status of some groups in temporary seki.
Or, one could forbid the placement of any stone without liberties. The consequences of this rule would be similar to reverse NZ rules.
Note that of these capture rules, NZ rules are most similar to traditional Go and have a minority following in the Go community; the other variant capture rules listed lead to a different sorts of games and are rarely-played novelties. We list all six possible capture rules here so that the alternatives to the traditional capture rule can shed light on what that rule actually says.