A pedantic question about superko

In Diagram 1, White has just played O, capturing a stone at X. Black recapturing at X would exactly repeat a previous position and violate the superko rule.

In Diagram 2, White has just played O, capturing a stone at X. Black recapturing at A would create a vertical mirror of a previous position.

In Diagram 3, you guessed it, White has just played O, capturing a stone at X. Black recapturing at A would create a diagonal mirror of a previous position.

In Diagram 4, White has just played O, capturing a stone at X. Black recapturing at A wouldn’t create either a vertical or diagonal mirror of a previous position, but the resulting position would be rotationally symmetrical to a previous one.

My question is: how does the superko rule treat Diagrams 2, 3, and 4? Or have I misunderstood the rule fundamentally?


Superko allows the moves (at A) in diagrams 2, 3, 4, provided that the resulting whole board position had not previously been created earlier. It doesn’t matter that the whole board is a mirror of a previous position.


In Diagram 4, suppose we’re playing and then I spin the board around while the opponent isn’t looking. Is that considered a form of cheating?


See, I wouldn’t have said so since the position is still “the same”. Although in your example the appearance of the position remains the same as well. This is why I made this thread, since I wondered whether superko was focused on the “reality” or the “appearance” of the position, if you catch my drift.

No, the situation on the board remains the same.

1 Like

I think the reality is that it does not matter which ko is taken. Since there are two ko’s at symmetrical positions, taking either could be argued to be allowed from a game perspective. If this somehow ever happened to appear in one of my games, I would take the illegal ko immediately and then spend all my effort to convince the opponent that it should not matter :stuck_out_tongue:

But I’m leaning towards agreeing with @yebellz that the coordinates are fixed under the specific rotation / reflection at the start of the game.

Exactly, so my opponent will take the “wrong” ko, and I made them do an illegal move.

It’s a bit like those chess players who claim a win after pointing out their own move was illegal and that their opponent failed to point it out on time (Carlsen - Inarkiev)


But isn’t there a basic ko rule as well as the super ko rule, or is that only Japanese like rules that have a ko rule? As in superko is supposed to prohibit ko anyway.

Actually maybe I don’t understand what’s being said/referred to

It does make musical chairs go a bit confusing.

Basic ko is a special case of any superko rule, so rulesets with superko usually do not add a simple ko rule as well.

1 Like

No, it would not be against the rules to rotate a rotationally symmetric board such that nothing essentially changes from the perspective of the opponent. However, it would be contrary to the rules to say that it impacts the superko bans (from the perspective of the opponent). You cannot claim to have tricked your opponent this way.

The superko rules apply to the absolute board position, but it is not tied to the literal physical representation. The board position is an abstract concept, and superko would still apply even if both players were playing only by memory by announcing moves aloud.

Imagine the board (with a rotationally symmetric position) is on a lazy Susan and was spun around an unknown number of times. Or imagine if the stones are all knocked off and the players reconstruct the position from memory. Neither of these should strategically affect any superko bans, since its the abstract history of the game that matters, not the physical representation with specific stones.


The same argument holds for forbidding to play on either position, since playing either way leads to a previous board position.

So is it saying because the position is symmetrical, if I swapped seats with the opponent say (or rotated the board, swapping seats seems more practical) then technically from that perspective capturing the other ko looks like it violates the superko rules.

You could just give every point a unique identifier/coordinates, which breaks all symmetry. I presume this is something you imagine should be in the superko rules if one wants to completely avoid ambiguity?

I think it’s only of practical relevance in some 9x9 positions, perhaps, and for working out algorithms for constructing the longest possible game of Go (which is what led me to this thought in the first place.)

I do have an opinion, which is that superko should treat the position independently of the observer, ie. repetition of mirrored or rotated positions should be forbidden. But this is an edge case, which is why I said it’s a pedantic question ^^

1 Like

I feel like the game itself shouldn’t care about the observer, but a game in progress should.

As in if someone replayed an AlphaGo game to me on a board but I was looking at it upside down on its side or from an opponents perspective to the one it was recorded, I wouldn’t want to treat it as a different game.

For a superko rule though, I wouldn’t want to treat it as though you can take any symmetries into account during game though since they’re not going to be symmetries of the board for a large fraction of the game.

So move 32 would be forbidden here because of a mirror symmetry (or move 16 in the second branch)?

In the third branch, black can’t stay alive in double ko because of an arbitrary board symmetry?

[Maybe I can flesh that example out into a symmetric position where black can kill the white groups on either side?]