Spight Rules

  Difficulty: Advanced   Keywords: Rules

I have proposed these two rules[2] which address the questions of endless repetition and stopping play at the end of the game.

  • A board play may not repeat a previous whole board position unless a pass has intervened since its last occurrence.
  • Play stops when the same player passes a second time in the same board position.

With area scoring play ends when play stops. With territory scoring play stops twice. After the first stop there is an encore in which a pass costs one point and each player makes the same number of plays (considering a pass a play). Also, passes before the encore do not count for stopping play in the encore. After the second stop play ends if each player has made the same number of plays. If not, the second player in the encore makes an obligatory pass and play ends.

/Discussion

-- Bill Spight


Feedback request

Bill: I am thinking of making a complete set of rules that people might actually play by. I would like to get some feedback on different rules for stopping play. While I think that my original idea, above, is adequate, there are other possibilities that people might like better. Here are the current alternatives I have in mind. Please comment below. Thanks. :-)

1) Play stops when the same player passes a second time in the same position. Play may resume with the passer to play, with ko bans lifted.

Or

2) Play stops when a player plays to a position where his opponent has passed before or passes in that position. Play may resume in turn, with ko bans lifted.

Or

3) Play stops with two consecutive passes. Play may resume in turn, with ko bans lifted, except as provided next. If a player has made the last pass in a particular position, his opponent may never make a board play that returns to that position.

Note for all alternatives: There may still be a ko or superko ban against returning to the current position.


Comments here, please.

The Count: I put my comments at the end.

Anonymous: This 3x4 example shows a game which ends (by Spight rules) in a way that might be considered unnatural.


{Explanatory discussion under construction.}

Spight rules are based upon the idea of evaluation. First, play should stop in a position that has a definite value (score). Second, the score of such a position may be determined by play with passes that have a value.

Using passes to evaluate positions is based upon the idea that if one player makes a play that gains a certain amount, and then the other player makes a play that gains the same amount, the result is the same as the value of the original position. If we treat a pass like a play and the value of a pass is the same as the value of best play (under these conditions) in a position, the resulting position after a sequence of best plays and passes with the second player passing last will have the same value as the original position.

[Diagram]
Half point  

By convention the framing stones are alive. We count the Black territory in the corner as 1/2 point. (OC, at the end of the game it will be worth 0 or 1.) The miai value of a move at a is also 1/2.

Suppose that the value of a pass is 1/2 point, that Black plays at a, and then White passes. Black gets 1 point in the corner but White gets 1/2 point for the pass, so the result is 1/2 point, the same as the original value of the corner.

[Diagram]
Ko  

Here the corner is worth 1/3 point, and the miai value of a play is also worth 1/3 point.

Suppose that passes are worth 1/3 point. Then Black may take the ko, White passes, Black fills the ko, and finally White passes. Black gets 1 point for the captured White stone and White gets 2/3 point for the two passes, for a result of 1/3 point, the same as the original value.

Note that this works only if White's first pass lifts the ko ban, just as a board play would. Otherwise Black is not obliged to fill the ko.

All of this, of course, depends upon the value of a pass and the value of a play being the same. If passes are worth zero, and plays are worth zero or less, correct play in this environment of passes will not change the value of the position. The value of the position will remain the same. Thus, we can say that the position has a score.

Of course, it may be possible to calculate a score for some non-terminal positions, but only some positions are such that that calculation would produce the same result, even if play were to continue. Only such positions have a constant value that coincides with the score. They are scorable.

By area scoring, if we treat each stone on the board as alive, we can count one point for each stone on the board and one point for each point of surrounded territory. If some of the stones are dead, the result of that procedure will not be the value of the position. Provision may be made to remove dead stones by agreement, but continuing play should also lead to their removal, to reach a position with a constant score.

In the next example, under some area rules the game may end in a position where the current result is different from the one if play were to continue.

[Diagram]
What result?  

This example is a variant of one of Ing's. White to play.

[Diagram]
Diagram 1  

B4 = pass.

By AGA rules and some others, White can now pass and win the game by two points.



This result has its defenders, but it is anomalous. The game has stopped in the middle of a ko fight. It is a kind of Moonshine seki.

Under Spight rules play continues. Black's pass has lifted his ko ban.

[Diagram]
Diagram 2  
[Diagram]
Diagram 3  
[Diagram]
Black wins  

Usually rules differences produce small score differences. Here the difference is 27 points.


Since Wilton Kee has shown how his rules apply to the 2x2, I am following suit.

[Diagram]
Greedy Black  

With area scoring, Black may get greedy and try to win by one point. (This is possible under some rules, I believe.)

[Diagram]
W6 = pass  
[Diagram]
B9 = pass  
[Diagram]
W2 = pass  

Note: If B1 were in the top right corner, White could pass and end the game favorably.

[Diagram]
B5 = pass  

Now Black is in a dilemma. No matter where he plays, White can pass and end the game favorably.

So Black loses if he passes at move 15. Instead,

[Diagram]
Black takes  
[Diagram]
White takes  
[Diagram]
Game ends  

W8, B9, W10 = pass.

Since neither player can make headway, they pass, ending the game as seki. (Note that Black loses by 2 under territory scoring.)


The Count: Because rules 1 and 2 don't say "never" and resumptions are possible, can't you have endless cycles? For rule 2, say there is a single stone ko. Black takes, White passes, Black passes. Game ends. Game resumes. White takes, Black passes, White passes. Game ends. Game resumes. At this point, the cycle take-pass-pass-take-pass-pass could go on forever with the game ending and resuming after every move. Have I got this right? Maybe play can resume only once.

Anyway, I think rule 3 is conceptually the easiest, especially if you phrase it, "No player may ever return the board to the position in which the game temporarily stopped."

How about this, though. On their turn, a player can make a board play, pass, or "lift" (lift all ko bans). A player cannot "lift" a second time in the same position; all other rules the same. Similar to rule 1, but conceptually easier? Maybe you would object to adding a new type of move, "lift". The thing is, I think the consecutive-pass end rule is very nice. In my opinion, a pass should imply a player has nothing left to do, and two passes should imply an infinite pass cycle is about to begin, so the game can safely end.

In addition you could apply your rules to no-pass go with prisoner return, because when a player makes the special "lift" move, the rule could be that they give their opponent one of their own stones instead of returning one of their opponents. Otherwise, a player may not have the luxary of passing if they have no prisoners.

By the way, I'm not at all convinced that passes should lift ko bans.

Bill: Many thanks to The Count for your perceptive comments. :-) As for endless cycles, they do need another rule to say when a stop is final. I wanted to allow non-final stops to let the players agree about dead stones without having to play everything out.

The Count: Okay, this was originally going to be feedback to your Japanese-style ruleset, but I realise the main thing I have to comment on are the ko rules as addressed here.

If it is not required that endless repetition is forbidden, the rule for stopping play can be derived from the simple principle "the game must not end with unresolved kos". Normally, two passes ends the game because they indicate both players have nothing left to do. When passes lift ko bans, the passes may not indicate this. If you have three passes in a row, because the first one lifted all ko bans, the last two must have been an indication that both players have nothing to do in a ko-ban free environment and so all kos must have been resolved; the game can stop. (In fact, it's overkill because you don't always need three passes, but whatever.) We can easily imagine a situation where endless repetition is allowed to occur by this rule. For instance, in a single stone ko the six-move sequence of capture-pass-pass-capture-pass-pass can go on indefinitely. Another example is if a molasses ko is left on the board at the end of the game. The endless cycle would be the four moves in the ko followed by a pass followed by the opposite four moves in the ko followed by a pass.

The point is that the principle "the game must not end with unresolved kos" tells us what the stopping condition should be and furthermore that we must allow endless repetition if we are to abide by this principle. In your discussion of the proposed Spight Rules, you give an example of Ing's and say

"...it is anomalous. The game has stopped in the middle of a ko fight."

referring to the result with normal superko rules. You have implied that the aim of Spight Rules is that play must never stop in the middle of a ko fight. At least, I hope that is what you intend, because I think this well-defined principle can be the only reasonable justification for adopting the rule. Therefore, you must use a stopping condition which allows endless repetition. You mentioned in reply to my earlier comments about temporary stops and final stops, but the point is that if there were a rule about a final stop, it could theoretically end the game in the middle of a ko fight.

Because you must allow endless repetition anyway, I think this rule works better with the basic ko rule (not superko). In fact, because I can see a rationale for how it all works, now I quite like the idea.

Bill: Thanks, again, Count, for your response. :-)

A couple of quick comments: The aim of Spight rules is to provide rules based upon evaluation, as scores are final values. This applies to ko fights, as winning the ko results in a better score than losing it. To stop in the midst of a ko fight means stopping in a position theoretically without a score, even if it is scorable by other rules. BTW, I would not call rules that stop play in the middle of a ko or superko fight normal. ;-)

To say that the game must not end with unresolved kos is problematic. When is a ko resolved? What, for instance, about double ko seki? What about 10,000 year kos and approach kos? Under area rules, what about sending two, returning one? One aim of these stopping rules is to say when a ko or superko is resolved, even when it remains on the board.

As for your comments about the Spight Japanese style rules, I am eager to hear them. Please put them at the end of that page. Thanks. :-)

The Count: I will!

The Count: Okay, I had a very specific definition in mind.

A: "The game can end only when the players have passed consecutively."

If passes lift ko bans, add this:

B: "There were no ko restrictions when these passes were made."

A is the normal idea in superko rules because it means that both players agree that there are no worthwhile moves left. If passes lift ko bans, A on its own will not ensure that the players agree there are no more worthwhile moves. With B, it is ensured. The rules don't state that the game ends because players don't have anything left to do, but I think that most would agree that this is the reasoning behind two passes ending the game. If passes lift ko bans, it is crucial to use B in combination with A.

(Edit: in hindsight, I don't think this "worthwhile moves" thing is precise enough. See here.)

So, you have said you want a ruleset which is as close to the theory of evaluation as possible. I don't really know anything about CGT, but the following is what I think it means. A position can be assigned a property which we call the score. In go, the score of a position depends also on forbidden positions. So if you try and apply CGT using only the position of stones on the board, you can run into trouble. Is this right? Assuming it is, then if the above conditions are met, a position's score is always definite at the end of the game. It cannot depend on ko restrictions, because we are not allowing them to exist. So, if I understand what you're trying to achieve, these conditions are exactly what you want. In addition, I would imagine that many other people agree with idea of "agreement there are no worthwhile moves left", and so it could become popular in its own right.

So, if the game ends after three passes, we can be sure that conditions A and B are met. Perhaps this is easier than applying A and B directly which means sometimes only two are required. It should be obvious anyway. How does this apply to the situations you mentioned? (Of course, you already know.) One player can force endless repetition in double-ko seki if the other won't give in. Again, in sending two, returning one, one player can force endless repetition and the corresponding result (draw, no result, or whatever). This isn't the case in traditional Japanese rules, but of course that is not important.

Not sure what the approach ko is about. Perhaps Takahashi (from the Takahashi-Segoe game) can argue the thousand-year ko is "unresolved". It was my mistake to talk about "resolving kos". I was trying to simply represent the principle of the rule, so you have an intuitive feeling for what the point of the rule is. I guess the intuitive idea that I really meant was that players agree there are no more worthwhile moves. Oh, you do mention it too. You say, "One aim ... is to say when a ko or superko is resolved." I say, "A ko is resolved if the game ends using A and B."

Okay, that's the bulk of it. There is more though.

Admittedly, I had previously thought this was the only possible idea, but I think there is another: the players need not agree that the position they reach is the final score, as long as one can prove it "should" be. We can think of a pass when there is no ko restriction as a player effectively saying "I think we should score now." So, we might use the following definition.

A' "The game can end only when a player passes[1] in a position with the same score as the last position that either passed[1] in."

[1] B' "The above passes are only counted if they were made when there were no ko restrictions." (Same as B really.)

So in double-ko seki, one player is saying "I think we should score now" and the other player, who disagrees, cannot find a new position for which they can say "No, we should score in this position." and so once you return to the original position, the first player says "I told you, this is the final position! I've had enough! Let's score!" Same for sending two, returing one.

This makes both double-ko seki and "sending two, returning one" a seki position, not an endless repetition situation. However, two sending-two-returning-one situations becomes endless repetition again. Hmm, how do traditional Japanese rules deal with this? One player can force endless repetition as soon as two of these positions arise, and so you can't tell the player it is seki and to stop playing there! I would have to say that the conceptual idea of this scheme is less straightforward. However, it might be perfect for your Japanese style rules. I guess it would be very suprising if it matched up exactly though.


Further reading at Spight rules II


Notes

[2] For their blind spots see repeated-pass rule.


Spight Rules last edited by RobertPauli on September 30, 2019 - 11:11
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
RecentChanges
StartingPoints
About
RandomPage
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Goproblems.com
Login / Prefs
Tools
Sensei's Library