Superko rules

I think I was able to do an infinite go game (even with the ko rule. Or maybe it’s a bug.)
I tried it in the analysis and it worked.

The infinite game goes like this:

Black: A1 C1 B2 G1 F2 E5 D6 E7
White: D1 F1 E2 A2 F5 E6 G6 F7
Black takes on ko 1
White takes on ko 2
Black takes on ko 3
White takes on ko 1
Black takes on ko 2
White takes on ko 3

Black takes on ko 1
White takes on ko 2
Black takes on ko 3
White takes on ko 1
Black takes on ko 2
White takes on ko 3

etc.

Is this supposed to be allowed?

It depends on the ruleset. Under Japanese rules this can indeed happen, hence games that feature a triple ko are declared void (no result).

Under Chinese rules there is the (positional) superko rule, which makes it forbidden to repeat a board position, and thus after going through each possible permutation of the ko’s being occupied by black or white, it becomes illegal to take back any of the ko’s.

2 Likes

Yep that’s the two families of rules on that, go check what’s the choice for the other rules :upside_down_face:

1 Like

You could check it yourself. Most rulesets use a form of superko. Only Japanese and Korean have a ko rule that forbids immediate recapture of a ko without having problems with repeated board positions.


I made some demo boards to show the differences, and to show there’s no bug on OGS:

Japanese (no superko)

Chinese (positional superko)

AGA (situational superko)

7 Likes

@yebellz has discussed this before in depth, iirc.

I also made a thread about whether superko takes into account positional symmetry (the received wisdom seemed to be that it doesn’t).

See also https://senseis.xmp.net/?Chosei, another endless situation in Japanese rules.

1 Like

So in that position is there any difference between situational and positional superko?

I did see sensei’s library mentioning a difference in this one https://senseis.xmp.net/?PositionalVsSituationalSuperko

I’m not sure what your question is? Yes there is a difference, since black is not allowed to “recapture” again in positional superko, since it repeats a position, even though the other player has their turn.

Well my question was just “Is it the case that, both positional and situational superko rules, treat the exact position you set up in the two demo boards the same”. I think the answer was “yes” in that they both forbid White playing A in the third ko at the same time. Maybe it gets more complicated after that, I don’t know.

So that’s why I made two demo boards, to see how OGS distinguished between the two superko types with the two rulesets AGA and Chinese. You’ve added those examples to the demo boards now anyway I guess. That or I just missed them before.

I guess it was just that I couldn’t see the difference or difference of implementation of the superkos from the triple ko position.

You must’ve missed them, since those second examples were there since the beginning :slight_smile:

1 Like

Definitely did, I think I was focused too much on the triple ko, and once I was far enough along the move tree I didn’t even see the second branch :slight_smile:

Hehe see, I didn’t have to check it myself. Thanks anyway for a nice resume.

“Superko” is the silliest term in Go.
Most rulesets simply have a rule that prohibits a move which repeats the whole position (with the same player to move). Ko is just a special case of the no-repetition rule.

Japanese rules didn’t adopt the no-repeated-position rule completely, just the simplest case, calling it “ko”. Maybe when Go was first introduced to Japan, and somebody taught the rules to the first Japanese Go player, the simplest case of repetition was used as an example; and the beginner thought that was the complete rule. We’ll probably never know, the beginning of the Japanese rules may be lost in the mists of history.

1 Like

I think there is a fair bit of speculation here. For one, as far as I can find we don’t know what the historical ko rule for Go even was. Even discovering whether area or territory scoring was used is quite hard with the limited ancient literature on go (the answer is actually neither: stone scoring was used, i.e. area scoring with a group tax and no points in seki).

Who knows, perhaps the Japanese ko rule was the actual ko rule: after a capture of a single stone, it is not allowed to immediately recapture. It sounds a lot easier than “it is forbidden to repeat a previous board position”, especially considering the very rare occasions where it actually would’ve mattered, it’s quite likely that the flaw in the naive rule only became apparent pretty late in the history (I mean, look at how long it took for them to realise black has a 7 point advantage)

6 Likes

Yeah it does seem that with how likely it is to have a ko shape appear on the board, on the edge and in the corner, in endgame and life and death, it wouldn’t be all that surprising to find just that case banned as part of the rules, the simplest of repetitions. Triple kos that actually matter and things like the send two return one, don’t seem to come up as often so it would be hard to imagine incorporating that into the ruleset nationwide/globally etc (and the send two return one is already sorted out by captures in territory scoring).

Actually with the rule disputes with Go Seigen that cropped up Rule disputes involving Go Seigen at Sensei's Library it wouldn’t surprise me if Japanese rules were just territory scoring + (rules for whatever came up that was needed to make territory scoring work).

1 Like

Basically I miss historical studies on go rules, and even better in some kind of scientific way. I just have my “shared ideas” (so not scientific at all) like the transmission from China to Japan, then the birth of the Japanese rule trying something more visual and hoping to codify exhaustively, then the superko a relatively new concept in the weiqi world.

Have you checked the link I posted above? It seems that the writer, Chen Zuyuan / 陈祖源, is a historian who wrote no less than 35 articles about historical Go, but since I can’t read Chinese it’s hard for me to find anything more than just this English essay.

1 Like

By the way, you need to be extremely careful and precise when talking about the ko rule in “Chinese rules”.

  • The nominal written rule in Chinese rules is positional superko, but…
  • The actual ko rule most often used in practice as far as I know is simple ko except sending-two-returning-one cycles is also forbidden. For example, if I recall right, you can find that in the rare case that a triple ko occurs in things like pro league matches, the game can be voided as it would be under the simple ko rule, instead of the players fighting it out under positional superko.
  • Online servers that claim to implement “Chinese rules” might vary in what they do. Some of them (for example OGS) go with the written rules of positional superko, despite this not matching real-life practice.

So the state of things is sorta messy. Don’t take it for granted what ko rule is or is not in effect for “Chinese rules”. :slight_smile:

6 Likes

Forgot to thank you. That’s the kind of things I was looking for.

1 Like

The problem is that, in Japan, Korea and China, “the written rule” is secondary. In practice, the referee’s decisions, based on historical precedents, tradition and the general consensus opinion of the professional players, is what matters, even if it contradicts the written rule, which works as the general “guide” but can be overridden if a rare, strange position appears.

This is something critical, that most western discussions completely miss, probably due to a “western thinking vs eastern thinking” difference. A phrase like “You could check it yourself” is actually sort of a trap: unless you research what the actual historical precedents and consensus of the Chinese Pro Association is for a certain rare repetition position, the written rule is not enough to clearly know how the Chinese rules would handle it (unless it happens to be one of the specific examples mentioned, like moonshine life).

As hexahedron pointed out, the written 2002 chinese rules say something like “in principle repetition of the whole board is forbidden, but for some positions like triple ko, eternal life, round robin ko etc the referees can void the game or apply other rulings depending on the specific cycle”. Triple ko (or any other infinite repetition during the “main game”, not just at the game end when everything is completely settled) = void game / draw is as far as I know the universally accepted result in China, Korea and Japan.

For endgame-appearing repetitions, “moonshine life dies” and “unbalanced repetitions cannot be played” (sending two returning one, sending three returning two etc where one side captures more stones in each repetition) seem to be used in the actual Chinese rules, and should be the most important and common precedents.

I have not idea how the Chinese would rule a “bent4 + double-ko-seki” for example. A triple ko would appear when one goes for the kill of the bent4 at the end of the game, unless superko is actually used in that case as it is for moonshine life. Does anyone know what the actual Chinese would do in that case? Apply superko, or void the game if killing the bent4 instead of seki matters? I doubt that there is any professional precedent, as double ko seki + bent4 in the same game sounds so incredibly rare…

An even more fun rule monster question is to take a bent4 + unremovable-ko-threat (seki) , so that the bent4 lives under chinese rules (as is shown in Life and Death under Chinese Rules). Now add in addition to that a double-ko-seki on the board. Does the bent4 still live? Or is it now a void game due to triple ko? Or maybe the bent4 now dies instead of living, in a similar way to how the “one sided infinite repetition” of moonshine-life means that the attacker wins? Unless there is a professional Chinese game precedent or a specific statement by the Chinese Association about this positions, it seems impossible to know what a referee would decide.

2 Likes

That’s true.
Instead of refering to a fixed writen rule and trying to apply it in a particular case, the focus is to find a compromise and kind of consensus as a usual way. Now this is something changing a bit too, use of writen law becoming less uncommon.

1 Like