Question about Eyes in Seki Under Japanese Rules

[YeGO]

A seki occurred in the bottom-right corner of this game on OGS:

https://online-go.com/game/25731373

I believe T1 and R3 are not counted as territory under Japanese rules (since the surrounding stones are in seki for sharing a dame point at Q1), however, I'm not sure about S5.

Do the black stones surrounding S5 "possess" dame and are considered seki as well?

[RobertJasiek]

Possessing dame is not a feature of all real-world Japanese rulesets but only of the Japanese 1989 Rules and its interpretations. Possessing dame is a property of a group of one or more than one strings. Therefore, for the Japanese 1989 Rules and its interpretations, R3 and S5 are treated alike (no territory of groups in-seki). For all real-world Japanese rulesets, there is no territory in sekis.

For most Japanese server rulesets or implied server scoring rules, there is territory in sekis. OGS is an exception because its reference to Japanese rules is ambiguous even as to whether it refers to real-world or server Japanese rulesets. If OGS should suggest an auto-score, it is ambiguous whether that or the reference to Japanese rules takes precedence. Therefore, there are two possible interpreations: 1) all of T1, R3, S5 are territory; 2) none of them is territory. Unless this ambiguity was dissolved by agreed context (tournament) rules before the game start, the game score can only be identified as either "derived from territory in seki" or "not derived from territory in seki". If necessary in a close game with fractional komi, the result might have to be declared as "either player wins and it is inconclusive which player wins".

[RobertJasiek]

Referee/moderator-1 made the mistake to assume that Katago would know the rules but was cute enough to admit his own insuffient understanding and calling a more experienced arbiter. Referee/moderator-2 made the mistake to assume that necessarily the contents of the Japanese 1989 Rules were valid for OGS, which does not have such an unequivocal reference to a particular (but equivocal) Japanese ruleset. His other mistake was to assume that this ruleset would use "group" in the meaning of set of a player's strings - the rules write "stones" when referring to in-seki. Speaking of group for this purpose requires interpretation.

[Vsotvep]

I'm moderator-2, just to clarify here. I was basing my comments initially on http://www.cs.cmu.edu/~wjh/go/rules/Japanese.html, which is a translation that uses the term "group". I then went to the original Japanese version of the 1989 rules, and found out it does not talk about "groups", but it does mention that stones can "possess" (有する) dame. But, this is still something that requires interpretation, since the exact meaning of 有する is not explained in the ruleset, as far as I can find.
Hence I'm not sure how you arrived at the conclusion that S5 is treated as an eye in seki under the 1989 ruleset.

About real-world / server rulesets, I was not aware that these exist. Are these written down somewhere (either in English or Japanese)?

Finally, about the rules OGS implements through auto-score: it unfortunately does not really implement any specific ruleset at the moment. Honestly, it is quite broken (but will be completely revised in a couple of months, I believe).

[YeGO]

Hi RobertJasiek,

Thank you for the replies.

Yes, I meant to ask specifically about how this corner would work out under the Japanese 1989 rules, regardless of this being a game on OGS.

I believe you are saying that Japanese 1989 rules would treat T1, R3, and S5 all as not territory. Please correct me if I'm wrong, since the rest of what I will say rests on this assumption.

However, how are the two distinct chains of the black stones in that corner defined as a single group under Japanese 1989 rules. I fully understand that they are the same group from a strategic perspective, but I'm confused about how the exact wording of the Japanese 1989 rules treats those as the same group. Why do the stones surrounding S5 "possess dame"?

[RobertJasiek]

The Japanese 1989 Rules in their English translation by James Davies use "group" in the meaning of "string".

I arrived to the conclusion that a "group" - in the meaning of "set of a player's strings" (for greater clarity, see my commentaries and interpretations of the Japanese 1989 Rules including my 2003 Rules and its main interpretation) - adjacent to at least one dame is in-seki as follows:

1. In particular, I also read the Nihon Kiin commentaries, see especially the second Nihon Kiin commentary "EXAMPLES OF CONFIRMATION OF LIFE AND DEATH", Life-and-Death Example 24, points a and c. There, it is implicitly clear that the meaning "set of a player's strings" is meant because the meaning "string" does not fit to what is written there.

2. I studied and researched in the Japanese 1989 Rules for 10 years to eventually write my so far final commentary http://home.snafu.de/jasiek/j1989c.html and interpretations http://home.snafu.de/jasiek/j2003.html and http://home.snafu.de/jasiek/j2003com.html .

Some Japanese rulesets are:

Official real-world rulesets:
- Japanese 1949 Rules
- Japanese 1989 Rules
- World Amateur Go Championship 1979 Rules
- World Amateur Go Championship 1980 Rules
- Verbal European-Japanese Rules

(Implied) go server rulesets (unless overridden by tournament rules):
- IGS-Japanese Rules http://home.snafu.de/jasiek/igsrules.html
- KGS-Japanese Rules (implied; with territory in sekis)
- OGS-Japanese Rules, i.e., the OGS reference to Japanese (style) rules on Sensei's Library
- WWGo Rules (without territory in sekis; URL inactive, my implied commentary http://home.snafu.de/jasiek/j1997.html)

See also http://home.snafu.de/jasiek/rules.html .

[RobertJasiek]

how are the two distinct chains of the black stones in that corner defined as a single group under Japanese 1989 rules. I fully understand that they are the same group from a strategic perspective, but I'm confused about how the exact wording of the Japanese 1989 rules treats those as the same group. Why do the stones surrounding S5 "possess dame"?

The Japanese 1989 Rules do not define this - they are ambiguous about this. Read my commentaries and interpretations:
http://home.snafu.de/jasiek/j1989c.html
http://home.snafu.de/jasiek/j2003.html
http://home.snafu.de/jasiek/j2003com.html
If, afterwards, you still do not understand it, then ask again.

[Harleqin]

The Japanese Rules are generally seen to be intent that “there are no points in seki”, i. e. none of the three enclosed points count.

Personally, I strongly suspect that this comes from a misunderstanding: an unwarranted generalization of a description to a proscription, but it seems firmly entrenched now.

[RobertJasiek]

OGS could clarify by a) referring to real-world Japanese rules, b) referring to a particular Japanese ruleset or c) declaring that there is no territory in sekis under Japanese (style) rules.

[lightvector]

However, how are the two distinct chains of the black stones in that corner defined as a single group under Japanese 1989 rules. I fully understand that they are the same group from a strategic perspective, but I'm confused about how the exact wording of the Japanese 1989 rules treats those as the same group. Why do the stones surrounding S5 "possess dame"?

The critical thing to understand about both the Japanese and Chinese rules is that culturally and historically, the written rules are unimportant. Culturally and historically, the rules were defined by consensus and agreement among the top professionals or professional associations, and by the precedents established or later overturned in the course of various disputes and rulings.** Trying to write down the rules and formalize them is more due to later Western influences, and is still considered unimportant and treated as an afterthought.

**This my understanding - people more in tune with the actual history and stance of the Asian professional Go world should feel free to correct me if I'm wrong.

This means that analyzing the exact wording of the rules and trying to apply them literally is mostly missing the point.

More important is to holistically look at the rules together with the commentary on the various examples and to extract out the *intent* of the rules - to build an intuition for how the actual professional Go associations and pros and tournament directors would rule on different cases based on the examples presented, because that, and not the written rules, is the final authority in practice. The commentary is a very important source here for inferring this intent. Basically, you want to fuzzily train your own brain's neural net toward the simplest possible intuitions and concepts consistent with the way all the rulings in all the commentary examples fall out, so that you now have an intuition for what the rules are "supposed" to be and how different cases would get adjudicated.

As Robert Jasiek points out, if you look at commentary example 24 at the bottom of https://www.cs.cmu.edu/~wjh/go/rules/Japanese.html, the commentary speaks of white's group still being in "seki" even after "a" is filled, due to the "dame" at "b". But white's two-eyed string of stones is not even adjacent to b. So it is apparent from this example that, regardless of the wording of the rules (which again, is unimportant), the intent is that for determining sekiness, dame-possession should be understood to extend through the whole group, not just the immediate string of stones.

So applied to the case in the OGS game (and ignoring the technicality that the players should be filling all the other dame, that they *can* safely fill), this should lead one to conclude that S5 is not a point.

I've read through Jasiek's 2003 formalization of the Japanese rules, and it's quite a marvel. As far as I can tell, it very much captures almost the simplest point in rules space that you would reach if you tried to fuzzily build an intuition consistent with all the examples and how that intuition would extend to further cases beyond those in the examples, and then tried to actually make it fully rigorous and precise.

[Bill Spight]

A seki occurred in the bottom-right corner of this game on OGS:

https://online-go.com/game/25731373

I believe T1 and R3 are not counted as territory under Japanese rules (since the surrounding stones are in seki for sharing a dame point at Q1), however, I'm not sure about S5.

Do the black stones surrounding S5 "possess" dame and are considered seki as well?

Yes. The only group that is not in seki and therefore possesses territory is the White group in the bottom left. Before scoring by Japanese '89 rules you must fill all the dame you are willing to.

As Robert and lightvector have pointed out, the J89 rules text is not clear on what it means to be in seki. But the intent is clear in the examples in the official commentary, and by history, although the J89 rules break with tradition. A string of stones (not defined in the J89 rules) does not have to be adjacent to a dame to have it. Below is my attempt at a clearer definition.

==== Article 8. Territory

After all dead stones, if any, have been removed, empty points that are surrounded by stones of the same player are called eye points, and any empty points that are not eye points are called neutral points. All stones that surround the same eye points belong to the same group. A group is called a seki group if it is adjacent to a neutral point, or if one or more of its stones is in atari. Eye points that are surrounded by a group that is not a seki group are territory. Territory belongs to the player whose stones surround it.

Comment: The current definition is unclear. First, it distinguishes eye points from neutral points, and later distinguishes some eye points from territory in an unclear way, talking about seki stones possessing neutral points that they are not connected to. I replace the idea of seki stones with that of a seki group, and include the idea of atari, to cover double ko seki and other strange seki.

See https://senseis.xmp.net/?SpightJapaneseStyleRules

[Vsotvep]

Another way to capture the same idea is to define "a stone possesses dame" as "a stone is part of a group that possesses dame", and defining a "group" as "a collection of stones of the same colour and eyes such that they are connected horizontally / vertically to one another", hence counting eyes as part of a group.

[jaeup]

==== Article 8. Territory

After all dead stones, if any, have been removed, empty points that are surrounded by stones of the same player are called eye points, and any empty points that are not eye points are called neutral points. All stones that surround the same eye points belong to the same group. A group is called a seki group if it is adjacent to a neutral point, or if one or more of its stones is in atari. Eye points that are surrounded by a group that is not a seki group are territory. Territory belongs to the player whose stones surround it.

I agree that after removing dead stones, defining seki is doable with simple words. However, in some cases, removing "dead stones" is not recommendable. (One example is a double ko seki. Another example is the case when someone forgot to capture dead stones inside a seki.)

When I tried my best to logically define seki stones leaving all dead stones on board, I came up with this sentence. (Assume live and dead stones are identified, and thus dames and eyes are also identified. Still, the task is not easy.)

The condition for a live Black stone to be a seki stone is as follows: Starting from the Black stone if a path to a dame exists only through i) "live Black stones" ii) "dead White stones each having a path to a Black eye only through dead White stones" and iii) "empty points".

Don't worry that you can't understand it in 5 minutes. Whenever I come to this part of my book, I check all sample figures I prepared three times to make sure that the definition works. (I tried to make it simpler, but anything less than this will create a non-intuitive counterexample.)

BTW, as was already discussed above, the original Japanese text only includes the words "stone", but while reading it, one feels that they implicitly assume the concept of "collection of stones" (which one may call a group) in a sense that all 20 Black "stones" of the above shape are collectively identified as seki stones.

[Bill Spight]

==== Article 8. Territory

After all dead stones, if any, have been removed, empty points that are surrounded by stones of the same player are called eye points, and any empty points that are not eye points are called neutral points. All stones that surround the same eye points belong to the same group. A group is called a seki group if it is adjacent to a neutral point, or if one or more of its stones is in atari. Eye points that are surrounded by a group that is not a seki group are territory. Territory belongs to the player whose stones surround it.

I agree that after removing dead stones, defining seki is doable with simple words. However, in some cases, removing "dead stones" is not recommendable. (One example is a double ko seki. Another example is the case when someone forgot to capture dead stones inside a seki.)

My Japanese-style rules do not make use of hypothetical play, which assumes correct play. Instead, they use an operational definition of life and death, which depends upon the agreement of the players (which could be mistaken, OC) or upon the result of actual play in one or more encore rounds of play (which could include mistakes in the play). Here is the rule about ending play.

Article 9.3. Ending the Game

If the players agree about dead stones at the end of a round of play, as provided in section 1, the game ends.

If a round of play ends in a position where a previous round ended, the game ends without agreement. All stones on the board are alive.

Since stones in double ko seki and the like are not dead by this definition, the definition of seki states: A group is called a seki group if it is adjacent to a neutral point, or if one or more of its stones is in atari. If the game ends without agreement, all stones on the board are alive. If a player could have captured a stone but forgot to do so, tough luck.

BTW, during agreement, each player identifies only their own stones as dead. If their opponent thinks that they have left out a dead stone, play may resume and the opponent may try to capture it.

When I tried my best to logically define seki stones leaving all dead stones on board, I came up with this sentence. (Assume live and dead stones are identified, and thus dames and eyes are also identified. Still, the task is not easy.)

The condition for a live Black stone to be a seki stone is as follows: Starting from the Black stone if a path to a dame exists only through i) "live Black stones" ii) "dead White stones each having a path to a Black eye only through dead White stones" and iii) "empty points".

Don't worry that you can't understand it in 5 minutes. Whenever I come to this part of my book, I check all sample figures I prepared three times to make sure that the definition works. (I tried to make it simpler, but anything less than this will create a non-intuitive counterexample.)

Defining life and death by agreement or actual play means that the definition is not ideal, but depends upon the actions of fallible humans. One may argue about whether a stone ought to be dead or alive, but operational definitions are scientifically impeccable.

BTW, as was already discussed above, the original Japanese text only includes the words "stone", but while reading it, one feels that they implicitly assume the concept of "collection of stones" (which one may call a group) in a sense that all 20 Black "stones" of the above shape are collectively identified as seki stones.

OC, ishi in Japanese can be singular or plural, and the Japanese rules do not include words like string, chain, dragon or group. To the extent to which that causes ambiguity or confusion, well, the Japanese can be inscrutable even to themselves.

[jaeup]

OC, ishi in Japanese can be singular or plural, and the Japanese rules do not include words like string, chain, dragon or group. To the extent to which that causes ambiguity or confusion, well, the Japanese can be inscrutable even to themselves.

Right. You understand the subtlety of the situation very well. Because of such ambiguity, its direct translation into English is simply impossible. The translator (James Davies, right?) ended up using the word "group", but I cannot agree with such a style change. The actual rule always talks about "stone(s)", such as "this stone is a seki stone". It never says like "this group is alive by seki".

Fortunately, the structure and practice of the Korean language allows one to make an almost perfect direct translation of the Japanese rule. Of course, unfortunately, after finishing the translation, I still cannot decide if a specific sentence talks about one stone each or a collection of stones.