AGA Rules

    Keywords: Rules
  • Specific features:
    • When passing, a stone must be handed over to the opponent. This stone becomes a prisoner.
    • White must pass last. This means that the game ends after 2 or 3 passes depending on who passes first.
  • Scoring method: Area
  • Counting method: Japanese (Chinese counting can also be used (producing the same result), but it is very rare to do so)
  • Superko: Yes
  • Komi: 7.5
  • Suicide: Forbidden
  • Points in seki: Count
  • Cost of moving in one's territory: 0 points
  • Life and death settled by: Game resumption
  • Illegal move: Is regarded as a pass
  • [ext] Full text in English
  • Used in: the USA
    • Nearly identical rules are used in Britain and in France

The AGA rules are the rules of Go adopted by the American Go Association.

The rules are intentionally formulated so that there is almost no difference whether area scoring or territory scoring is used[1]. This is made possible by requiring white to make the last move (white must make the last play) and incorporating "pass stones". This means that if white passes first, he or she must pass again after black, handing over a second pass stone. Eyes in seki situations are counted as territory in territory scoring and are part of the area in area scoring.

They prohibit suicide.

(They are therefore quite similar to Chinese rules.)

They use the situational superko rule: it is illegal to play in such a way as to repeat a previous board position from the game, with the same player to play.[2]

In theory the rules allow free placement of handicap stones, but in practice the traditional Japanese placement is usually used. This is mentioned in the commentary at [ext]

The complete text of the rules can be found here: [ext]

Notice that the 1991 date in the above document is wrong: the rules were changed in August 2004, with komi set to 7.5 rather than 5.5, effective 2005; see point 9 in [ext]

[ext] This commentary page is an important companion document without which some rules cannot be fully understood. The commentary document is already referenced elsewhere on this page, but because of its importance I'm adding another link that is physically closer to the link to the rules text. -- herzbube

The French rules and the British Go Association rules are essentially the same as AGA rules.

[1] Territory counting (with pass stones) is used by default unless the players agree before starting play to another counting method. The Mathematics of Scoring shows the equivalence algebraically at the end of the page.

Deacon John The Mathematics of Scoring shows that territory counting (with pass stones) is mathematically equivalent to area counting. Area scoring (with pass stones) can not be mathematically equivalent to territory scoring, because area scoring and territory scoring do not always give the same result, even in commonly occurring situations. The Gun Eight seki pattern provides a nice example of the difference between area scoring and territory scoring in a commonly occurring situation.

[2] Robert Jasiek: Many readers of the rules text read it as if it were meant to describe situational superko (but the major author of the rules, Terry Benson?, insists that [ext] natural situational superko was meant).

Willemien explanation about natural situational superko with example (I found editing Robert Jasiek's post inappropriate, maybe he disagrees with the explanation given)


According to [ext] the BGA has decided to adopt the AGA rules in october 2007. The document mentions that the board is proposing to get this change accepted (or denied) at the 2008 Annual General Meeting. Does anyone know if the 2008 AGM has already happened, and what the result was if it did? --Herman Hiddema

RobertJasiek: The AGM has taken place during the 2008 British Go Congress. The motion to adopt the AGA Rules has been accepted unanimously.

Herman Hiddema: Ok, thanks! I couldn't find anything about the AGM on the BGA website :-)

Strongeye: The AGA rules specify territory as: "Those empty points on the board which are entirely surrounded by live stones of a single color are considered the territory of the player of that color." But neither the rules, nor the commentary text, specify if seki positions are a) alive or dead, or b) eyes in seki strings count as territory. A direct and literal interpretation as far as I can tell would mean that they do, but it isn't explicit in this. Can anyone clarify? Thanks.

  • Strongeye: ok, so I have found clarification for this [fake URL removed] where it says "The rules count surrounded points in seki, but not the 'neutral' points."
  • BuggyMind: We don't need to appeal to fake URLs. Third sentence of rule 9: Any stones which the players agree could not escape capture if the game continued, but which have not yet been captured and removed, are termed dead stones. And according to rule 10, if the players can't agree on the status of stones, all stones on the board after both players pass twice in succession are alive. So seki is alive. Rule 12, Territory: Those empty points on the board which are entirely surrounded by live stones of a single color are considered the territory of the player of that color. Therefore, eyes in seki are territory.
  • BuggyMind: Oh, and I found the clarification I think you meant, at [ext]

Anon Rule 10 states that 'If the players disagree about the status of a group of stones left on the board after both have passed, play is resumed, with the opponent of the last player to pass having the move. <snip>' In AGA rules White must always pass last, so why do the rules not state that Black moves first in the case of resumption?

Pledger: Under AGA rules, the first two consecutive passes end alternation, and the game moves to the agreement phase. The rule that "White must make the last move" is not applied until after the agreement phase. Therefore, if Black passed last before the agreement phase, the players wait to see if there will be resumed play. If afterwards, resumed play or not, Black has still made the last pass, White will then make an additional pass. See Equivalence Scoring example for a demonstration of the rule. (comment edited by author)

Anon Thank you for that answer, in practice I have always seen the third stone given over automatically and then the agreement phase begun. That it should be as you suggest is totally unclear from reading the AGA rules themselves. It strikes me as something that ought to be properly clarified in the rules.

A discussion on handicaps at OGS: [ext]

Anon In the 'recommended' part of rule 10 It says "At any point, a player may resume play rather than continuing to indicate dead groups or passing". Perhaps it is better to add ' but may not cause to live any stones that they did not disagree on during the stone-touching bit' or words to that effect.

Flawed agreement procedure

The agreement procedure of the 1991 AGA rules is as follows, quote:

It is recommended, particularly if the players do not share a common language, that the following procedure be used to determine agreement on the status of groups. After two consecutive passes, the next player touches each connected string of opposing stones on the board which he or she believes to be dead. If the opponent disagrees, he or she also touches the same string. When a player is done indicating groups he or she believes are dead, he or she passes, passing a stone to the opponent as usual, and the opponent follows the same procedure. At any point, a player may resume play rather than continuing to indicate dead groups or passing. If both players pass and there was no disagreement indicated, the game is over, and all groups which the players have indicated as dead are removed from the board. If they both pass while a disagreement still exists, all stones remaining on the board are alive, and the board is counted as it stands. (The burden is thus effectively on the player who would be disadvantaged by such a result to resume play in the event of a disagreement.

What's wrong? Watch White:

Last two plays  

After Black took the last point, White threw in a stone. Black, assuming that she wants to avoid the extra pass, shrugs and passes. White passes too, stopping the game and starting the agreement procedure.

Black claims that the single white stone is dead, to which White does not object. Done, Black passes. Now White surprisingly claims that all black stones are dead, to which Black, of course, disagrees. He expects her to extend up or down, to which he would answer on the other side, but she surprises him again:
she passes!

What's that? That's not only four consecutive passes, that's also no agreement. The board therefore is counted as is, letting White win with 13 - 7 = 6 points.

Nonsense, of course. How can we fix that? My suggestion is to force the player whose claim was rejected to either play (onto the board) or lose.

On the occasion, let's cure another flaw: a pass could forestall a rejection. Just as we benefit from the digit zero, an explicit sign to agree to a claim is needed. I suggest to use what formerly was the rejection, and to replace that by simply starting the opponent's clock (or to tip onto the board if no clock is used).

Robert Pauli

Fixed agreement procedure for the virtual world

The recommended agreement procedure not only is useful if players don't share a common language, it also fits perfectly to the virtual world. Here's how I would apply it:

A double (not a triple) pass starts the procedure. Whoever passed first is the claiming player, called he, and whoever passed last is the confirming player, called she.

To provide a visual clue about what and who, his pass button is replaced by the end button and hers by the disagree button. These buttons are exchanged if player's roles reverse. Both pass buttons are restored on procedure exit.

He can choose between three options: claim, end, and play, she between two: agree (new) and disagree. The new option is necessary because I want them to alternate. Since she moved last, he starts. They alternately pick one option with a single click:

  • To claim that her stones are dead, he clicks on an unmarked opposing stone. That marks it and all stones connected to it with minus (first stroke).
  • To agree that her stones are dead, she clicks on any stone he just claimed. That changes those marks from minus to plus (second stroke).
  • To disagree, she clicks on her disagree button. That clears all marks, exits the procedure, and disables his pass button for once unless he claimed nothing in the last procedure.
  • To end claiming, he clicks on his end button. Besides adding a pass to the game record, that either reverses roles or ends the game in agreement, depending on being the third or fourth consecutive pass in the game record.
  • To resume play, he clicks on an unbanned empty intersection. Besides adding that play to the board and the game record, that clears all marks and exits the procedure.

All clicks onto the board that make no sense are rejected, e.g. clicking on a stone belonging to oneself or one already claimed, or clicking on an empty intersection that is banned by the rules.

If clocks also tick in the agreement procedure, switched by each click, that would be it. If not, however, they need some way to overcome a dormancy, i.e. getting no response. I suggest to allow them to apply their play or disagree option out of turn. However, if he hasn't yet claimed anything when she disagrees, he has to be allowed to resume with a pass, so in this case his pass button is not disabled for once!

All that easily translates to the real world, e.g. to disagree, she simply starts his clock by punching hers (or, without clocks, tips onto the board, as if to say: no, they stay). Note that there is no resume button for him because I do not want him to have to punch her clock in the real world.

The play option might look superfluous after the disagree option became play-forcing (to remove the flaw explained above), but if he notices that the double pass was premature (e.g. that he overlooked a single-sided dame in his favor), I want him to play, not to claim that all her stones are dead. Also note that a game end despite disagreeing about dead stones no longer is possible.

By the way, the AGA rules should clarify that only indications made after the last double pass matter, and that a triple pass does not start an agreement procedure. And their tournament regulations should clarify clock usage in the agreement procedure. The 2014 version is not clear about that, quote:

A game is over when both players pass in succession [. . .] and agree regarding the life and death of all groups.

and later

Players may stop both clocks only under the following circumstances: [. . .] e. Game's end; [. . .]

If game over is the same as game end, clocks would also have to tick in the agreement procedure, but I don't think that that is the case, unfortunately.

With the additional virtual option delay for not responding, here's all that can happen (besides both falling asleep) in [ext] EBNF:

Procedure  = {claim agree} (Resume | end Procedure2)
Procedure2 = {claim agree} (Resume | end)
Resume     = play
           | claim disagree
           | claim delay play
           | delay disagree

At last let's listen to Harry M. Warner (Warner Bros. Pictures, 1925):

[ext] Who the hell wants to hear actors talk?

Exactly, it should be non-verbal, it should be silent.

Robert Pauli

AGA Rules last edited by on March 31, 2022 - 02:39
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Login / Prefs
Sensei's Library