[Rare] Seki that doesn't look seki

shape that rarely seen


Isn’t it just a seki with one eye each, after capturing the two black stones and connecting up?

Capturing those two stones is definitely a good thing to do, though, it gives two captures without losing any points (since it’s territory in seki).


I can imagine white didn’t expect that to be seki during the game.


Black has no way to save those two stones, so why isn’t the corner just five points for white?

1 Like

Because Japanese rules say: “no points in seki”.

If the game was played under Chinese rules then you would be right, White would have 3 points of territory and 2 points of prisoners in that corner; plus an extra point of territory at E9 that shouldn’t be forgotten.

@Yutori_Sedai In the game, at move 66 it was still time for White to prevent the seki:

White played at A which was unnecessary, and that gave you the opportunity to make an eye in the corner and turn the whole thing into seki, sweeping all the territory away from White.

But if White had played at 1 instead of A, then there would be no seki and all the territory would still be White.


Well, I will refrain from arguing with Japanese rules.


I find this “no points in seki” rule the most quirky thing about Japanese rules. It feels unfair to me.
Also, it complicates scoring a lot, because this rule means that a scoring algorithm can’t just flood-fill after the players marked dead stones. The algorithm needs to recognize seki to score games correctly, which is a much more difficult task than simply flood-filling.

I would be in favour of dropping this rule from territory scoring, deviating from Japanese scoring rules to simplify the rules in the Netherlands, and perhaps other European countries that use Japanese rules now (all except the UK and France, I think?). Maybe call those simplified Japanese rules EGF rules?
But I don’t really know if dropping that rule would lead to other issues with scoring.


It ultimately gives three points to White, by capturing those two Black stones, and then another stone that is forced to throw in, in order to preserve the seki.

More specifically, it goes beyond just the concept of “eyes / territory in seki do not count” (under Japanese rules) and also applies the concept that dead stones in the eye of a seki position are not removed/counted for scoring.

I talk about this sort of situation here:


Further, it can be quite tricky to clearly recognize what really counts as a seki position. The rules are worded in a somewhat ambiguous way, and one needs to infer the spirit of the intended interpretation by looking at example positions in the commentary.


There have been many attempts and proposals towards simplifying the Japanese rules. See, for example the list given in Japanese Rules at Sensei's Library, under the heading “Other Japanese Style Rule Sets”. Some proposals involve removing the need to distinguish seki, which I think only simplifies things (at only the cost of deviating from tradition).

It seems that EGF allows a choice of rules for tournament play (see EGF General Tournament Rules):

  1. Ing Area scoring rules
  2. AGA-like Area scoring rules (it seems to allude to the concept of using pass stones such that territory counting is made equivalent to area counting).
  3. “Territory Scoring with Verbal European-Japanese Rules”, which requires “Arbitration decides interpretation disagreements”.

I think the last one aims to approximate the Japanese rules, but within the limits of the interpretation and arbitration abilities of those that are managing the tournaments.


As with most things related to go rules and their centuries of development, the way it is done currently is not just coincidence. This rule has important purpose, without it territory scoring would not be stable (and coherent with area scoring).

Correct scoring needs a final position with an unambiguous score, and sekis are a unique threat to that. Some (like a double ko seki) have multiple stable (possibly scorable) states, between which it can be pushed around. This is not necessarily a problem - as long as these states all have the same score. But this is where territory scoring start to have problems.

Under area scoring there is no scorewise difference between filling an intersection (like a ko mouth) or leaving it open and surrounded by live stones. Thus multi-state sekis usually still have - or can be stabilized to - an unequivocal area score. Under territory scoring, however, filling worth less than surrounding, which reduces possible stabilizing moves as they may lose points. Thus - if territory scoring would include sekis - some positions would have a stable area score, but undefined (oscillating) territory score.

Mannenko is probably the most common example where not excluding sekis would break scoring, see here.

If you are really averse to this rule, the simplest way around it is to have a pass-stone based (areafying) playout phase instead.



Could that issue be solved by not counting that 1-1 point as territory, because one of the surrounding stones is in atari (and black capturing that white stone only gives 1 point for the capture of the white stone)?

For example in this situation also:

… I think it would be reasonable to not count C19 as territory for black.

Adding this scoring rule would make scoring slightly more complicated than just flood-filling, but still not as complicated as detecting seki.

I think the addition of this rule would also ensure a stable score to a double ko position like this:
… where only black’s eye would be scored as territory.

Edit: Sorry, I didn’t check that last example carefully. It’s not a double ko.

1 Like

W stone is in atari and dead there. The problem is with the other state, when black has the ko. A black ko stone is alive, despite being in atari, because of the unique one-sided fillable nature of mannenko.


I think in that particular one, where black has an eye and white doesn’t, white is dead.


You mean this situation?

I think the problem here is not so really about how to score a seki. I’d say this situation is not quite settled unless black connects.

But if both players pass and go to scoring like this, I think the players should not mark any dead stones and neither player should get any points (similar to my 2nd example above).

1 Like

In that phase B has an intersection fully surrounded by live stones, so he could claim a point of territory.

Maybe you meant adding a special atari rule that takes away territory surrounded by alive but in-atari stones? That sounds like a dangerous ad-hoc thing, and I’m not sure how it would be better than excluding sekis (which gives a general solution to the whole problem, including other multi-state issues).

Btw, I see this rule simply as you need to have complete control to claim territory, which means pass-aliveable regions (LJRG by Pauli should also be mentioned).

Yes, that’s what I propose. The potential benefit being that detecting atari is much simpler for an algorithm than detecting seki. And I think this rule would also be less ad-hoc than the “no points in seki” rule.

But I’m no expert on rules beast positions, so I don’t know the potential danger of replacing the “no points in seki” rule by a “no points when any stones in the boundary are in atari” rule.

But suppose the opponent has some totally dead eyeless stones in my territory, with a ko, in which my ko stone happen to be in atari at the moment. Do I not have territory? Or consider a double ko death…

Go has a strange and fascinating completeness, that allows counterexamples for almost everything but the most rock-solid rule ideas.


Before applying the atari rule, players should mark dead stones and agree on that (implying that those stones are removed before scoring). This is the same as how games are normally finished and scored by Japanese rules (and I think the process is basically the same with Chinese/AGA/BGA/FFG rules).

Counting happens after dead stones are removed. Only then the flood-filling algorithm kicks in, with exception for areas excluded by the atari-rule if my proposal would be used.

If need be, the atari-rule could be limited to areas of only 1 intersection and only a single boundary stone being in atari, i.e. a ko shape. This would make the algorithm to apply this rule even simpler.


For Japanese rules you only remove dead stones from territory, which is not clear before seki check. If you start by removing all dead stones from everywhere, wouldn’t anti-sekis (where neither side is alive) or unsettled positions greatly favor the side who surrounds that area?

Highly specific rules remind me of 1949 Japanese, which iirc even had an extra rule about mannenko (forcing the player who can make seki to do so). But (side-) effects tend to be hard to verify, and I’m not sure if the atari rule would truly solve the underlying theoretical problem with all multi-state sekis (like current rules do).

1 Like