Odd Cases 🤔 in the Japanese Rules

General Go Discussion
122

Based on what I’ve found when having the chance of asking strong players and a few actual pros about their rules opinions, and what I have read online from rules experts, I agree with your main point that the 1989 Japanese rules were written not as “to have a well-written clear-cut ruleset by which to rule any position, following the specified procedure to the letter”, but more like to give some “general guiding principles explaining all the previous special rulings, so that the principles can serve if new strange positions appear, to create new precedents that are nevertheless coherent with the general principles found to hold in all particular precedents”. This was mainly because of many requests (both from inside Japan and outside Japan) to clarify the rules themselves, as previous written Japanese rules were basically just a long list of precedents by example, without any kind of “general rule” or “logic” explicitly within the rules.

That is, if a new rule dispute occurs, and the precise way to interpret the current written rules is ambiguous or ends up behaving in a “pathological” way, pros actually believe that the referees and rule-makers would have to set a new precedent, and would have the final decision. As far as I know, new “rule beasts” from 1989 are all theoretically constructed: no actual dispute which does not have a clear precedent established in the 1989 rules (like, say, a pseudoseki dispute could be) has occurred in Japanese Professional play, to the best of my knowledge.

There is a very peculiar case in the “Torazu Sanmoku” and similar positions, where in fact the new “general principles” and rationale of 1989 could not explain the historical Shuwa ruling of “3 points” for that position (that ruling follows logically from many other territory rulesets which solve disputes in a similar way to stone counting / area scoring, but not directly from a “purely Japanese territory ruleset”). Since they found that particular example to be the only one of the historical precedents which did not fit with the new general principles, they changed the precedent for that one in 1989.

But going back to your argument, even within the internal logic of the new “guiding principles” specified in the rules, I think that string by string determination is important. Basically the text of all the examples does this (sometimes it omits mentioning some string of stones, like in the example 6, but typically focusing on each string by itself, even when there are single-stone strings). This is crucially important in fact for example 25 (double ko seki) as I said: realistically, we know that “double ko seki behaves as seki and does not affect life and death outside of itself”, that is the Japanese precedent basically. Now, the REASON why that is so, according to the 1989 rules general principles, is that when considering the single-stone-strings (the ko stones) they are DEAD, because they are captured and recaptured infinitely, no permanent stone is established. While the large string of stones in each group is alive as it is uncapturable. This difference is crucial because it is what makes the double ko end up as SEKI, and not just normal life (that difference can change the result of a game, as there is no territory in seki). And I would argue that considering “in example 6, once you show that the 4 stones are alive then it doesn’t matter if the single white stone is dead, we can consider all alive” (if for example the eternal life cycle were the only viable sequence), then that would not be too different from saying “in example 25, once we know that the large groups are alive that is enough, we can count everything as alive including those single stones”, which would be wrong (because then by the seki-definition, it would imply that it is not a seki, as there is no dame between the group of live stones).

2 Likes
123

I don’t want to get into what “string by string determination” is supposed to mean because this isn’t a term of the Japanese Rules. But let’s say that I don’t disagree. More importantly though, I want to be clear on what I said:

I’m not saying that the 1 stone doesn’t need to be considered separately, just that in this situation, if that White stone is dead by itself, the 5 white stones as a whole would be alive. In this position that is because even if the 1 white stone is captured, new uncapturable white stones could be placed – meaning that the 1 White stone is not actually dead. This position is different from the double-ko seki because in that position there is no option to place new uncapturable stones.

As explained above, I do not think my logic is like saying "in example 25, once we know that the large groups are alive that is enough, we can count everything as alive including those single stones.” I never suggested that we ignore the definition of living stones. In Example 6 a new uncapturable stone can be placed while in Example 25 a new stone could not be placed.

After talking through this more, your Edit up above reminds me of a point from previous discussions. You state that in Example 6 White can show that the single stone is alive by letting the corner die when Black plays the eternal life. Something like:
image

However, this is inconsistent with the so-called “guiding principles” of the Japanese Rules. If the living status of the 1 White stone is being considered then Black would not play the eternal life because that is an attempt to capture the 4 white stones and it shows nothing about the 1 white stone. If the players were trying to show the living status of the 1 White stone then Black would attempt to capture the 1 stone, they would not play eternal life sequence. The diagram would be something like this diagram. Even though the 1 stone can be captured, it can place new stones that, even though they can be captured, can eventually place a stone that cannot be captured.

image

124

Ah right, this is show in Example 11 as well. The players are attempting to show the Life & Death status of the 7 White stones (白七子) in the upper diagram. This is the reason why White 4 is a pass (パス) for the ko in the upper diagram and not a pass for the ko in the lower diagram.

image

125

There is a very peculiar case in the “Torazu Sanmoku” and similar positions, where in fact the new “general principles” and rationale of 1989 could not explain the historical Shuwa ruling of “3 points” for that position (that ruling follows logically from many other territory rulesets which solve disputes in a similar way to stone counting / area scoring, but not directly from a “purely Japanese territory ruleset”).

I haven’t tried working out how thoroughly the “enable” part
goes against the following, but I think it’s worth pointing out that
3 points without capturing ​ would also be the result of scoring the corner as
all 5 stones in that corner are dead and there is no territory in that corner
(antiseki).

1 Like
126

A “normal antiseki” (where both groups are dead, as explained in the Japanese rules commentary) is zero points though (just as normal seki). Dead stones are NOT removed from the board in Japanese rules, if they are not inside the opponents territory. But I see your point: in a hypothetical ruleset where both groups were actually REMOVED as prisoners (not sure how such a ruleset would treat a normal seki? XD), and the underlying empty area left as “neutral”, the score happens to match the game-theory / chinese rules / stone counting / shuwa verdict of “3 points without capture”.

1 Like
127

Well,
all of the stones in most sekis ​ and ​ most of the stones in the remaining sekis I’m aware of
are alive, so the only consequence I see that having for normal sekis is, even if
the rules still don’t allow territory from them, it would be no longer necessary
to capture dead stones in sekis to get the points from those dead stones.
(such as in the seki outcome at ​ Walkie Talkie Eight In The Corner at Sensei's Library )

128

Even with that modified rule, it would sometimes still be necessary to capture them, if that enables even more captures, as in Diagram 17 in the Japanese rules.

Also that proposed rule totally changes the situation in “antisekis resulting from wrong play”, such as that in life and death example 9 (if the large black group were removed, black loses lots of points, but the rule text says that if black failed to properly defend it simply becomes seki, 0 points). Those might be “unimportant” though.

I cannot think right now of any terribly wrong consequence of your proposal to remove dead stones in seki for extra points (while leaving the 0 points of territory in place), but there might be some position where it is essential. Furthermore I think that such a hypothetical rule would solve the particular problem posed here, as the position in the thousand year ko becomes completely stable and final, even with infinite ko threats, if white’s ko stone is simply given to white as captured for the final counting (because it is ruled dead, unlike the corresponding black stone when black takes ko). Maybe @jannn knows of a more “normal” situation (not involving a player who failed to defend a hanging ko and such kind of antisekis) where allowing to remove stones in seki as extra prisoners (but keeping 0 territory points) gives a totally non-Japanese result.

129

Here is an odd case, shared by James Davies in 1996 (source):
image
Black to play, no komi. What’s the result?

If black adds a move to capture the white stones, both sides have 28 points, and the result is jigo.
image

Black would like to avoid adding a move, to gain a point and win the game.
So black considers what would happen if he passes and white plays first inside:
image
White can exchange 1 for 2, but after that it is clear that black is safe. The triangle is a real eye for black because of the double ko. Black has two eyes, white has one eye, so white is dead. Indeed this is true even under the special ko-rules used for status determination after the game is finished, so both sides can pass here and the white stones can be removed without further moves. Black wins by one point.

However, white had a stronger response! Instead of exchanging 1-2 above, she should pass immediately, ending the game. Then she can demonstrate that the black stones are dead as follows:
image
The exchange 1-2 is still forcing like before. But now when white captures at 3, black is not allowed to recapture 1 without first passing for that ko. Black has no way to stop white from capturing at A with the next move, and then at B with the move after that.

The conclusion is that black cannot pass in the original position, but must add a move and settle for jigo.

What makes this an odd case (in my opinion) is that:

  • In actual play, black is clearly alive in double ko.
  • Even though the 1-2 exchange is always the obvious start of the sequence, it makes a difference if it is played before or after the game end. If white makes this exchange before ending the game, then suddenly the black group is alive without needing an extra move.

Left board: black is dead. Right board: black is alive.
image image

7 Likes
130

I guess Japanese rules just has these quirks where there will be stones on the board that can’t be captured in actual play but are declared to be dead because of the special rules for the determination phase.

image
image896×480 15.5 KB

though of course your above example is a bit weirder where there’s a ko left on the board that might matter.

1 Like
131

That black group (“triple ko with an eye on one side”) being ruled dead is an intended consequence of the special Japanese status confirmation rules. There’s various ways to justify why it “should” be dead - for instance it also dies under superko rules :slight_smile:

However, it seems quite likely that the situation I talked about above is an unintended consequence of the new rules. Prior to the 1989 formalization, I see no logical reason that the two positions at the end of my post (left board vs right board) should be treated differently from each other.

2 Likes
132

Moonshine life is also a very weird case

image

In normal play, Black cannot capture the White stones on the left, since White has an infinite source of ko threats from the double-ko seki on the right.

The Japanese rules specifically mention this sort of situation repeatedly (as it is, and other very similar shapes) in the commentary and clearly state that the White stones on the left should be dead.

However, without the commentary clarifying that this should be case, a straightforward reading of just the rules articles seems to allow White to prevent capture, even in the hypothetical play (for life and death determination after passing) with different ko rules.

If Black takes the ko on the left to try to capture, White should take a ko on the right, which forces Black to take the other ko, but then White should pass in order to retake that last ko mentioned, which forces Black to pass in order to retake the ko that White took. This forces an endless cycle of taking kos and passing in order to lift ko bans, in the double-ko seki position (see Odd Cases 🤔 in the Japanese Rules - #12 by yebellz), which means that Black never gets the opportunity to complete the capture on the left. Despite this argument existing in principle, the commentary of the Japanese rules make clear that the articles should not be applied in this way.

5 Likes
133

It’s hard to know what is an intended consequence vs an unintended consequence, because I still don’t feel like I have the best literal version or interpretation of the rules available to me.

The examples show how it should be applied but they don’t explain explicitly the rules that should be applied as per

I’m not sure which part of the commentary says you shouldn’t apply it this way. Maybe the commentary on Article 7 here Article 1. The game of go but it doesn’t explicitly address what should happen if one has the forced endless cycle of passing to lift ko bans. I think when it’s ambiguous and there’s multiple reasonable continuations of how the rule could be applied it seems incomplete.

[
( I don’t understand this comment there)

Note that if White 7 was a pass for the ko 6/11 on the right, Black could have played 8 at 12 and captured the white group on the right. White could still not capture on the left and that Black group would remain alive.

]

The Japanese rules also always feel like an iterative approximation to some ruleset, but we’re not really sure what that ruleset is, until it gets clarified by professional play.

4 Likes
134

This is an old dilemma that was also discussed on L19 (here for example), and may be a simple misunderstanding of the pass for ko rule. The rule forbids recapturing in a ko that the player have not passed for, but it doesn’t clearly require a new pass for each later recapture in the same ko.

This is definitely an unclear point in the rules that need official clarification, without which we can only speculate. But the commentary on 7.2 gives some hints:

When the game stops, ko recapturing also stops

If a player whose stone has been captured in a ko has passed for that particular ko, the situation for that ko is the same as if the game had been resumed: the player may now capture in that ko again

Which seem to suggest that ko recapturing “stops” once, at the normal game stop (and not at each ko capture by the opponent), and passing for a ko “resumes” the game wrt that particular ko.

And OC, if passing for a ko is only needed once, there is no endless cycle in a double ko seki (as it will revert to a normal double ko seki after one cycle).

2 Likes
135

I think it is 1).
It is certainly not blacks territory, as all the black stones are dead. But without the white stone (which is dead) it also isn’t whites territory (as black would be alive in that case). So this stone is part of the surrounding stones. In other words white does not surround this with living stones. Neither player surrounds this region with living stones, so this is neither players territory. Dead stones are only removed from territory. So this stones are not removed, as they are not inside some territory.

This of course hinges on the fact what „surrounding“ exactly means, this definition is sadly not spelled out by the rules.

(maybe this argument is somewhere in this thread, but I couldn’t find it)

136

There is an alternative approach for alive, territory and stone removal in territory scoring. This is NOT how current (J89) Japanese rules work (some others do), but is probably close to what they aim at, including seki, unsettled etc.

Instead of going from L/D → dame → seki → finally territory and stone removal, it can also work backwards. Territory are surrounded areas that can (with further imagined play starting with opponent) be cleaned up and taken pass-alive control of (without enabling compensation elsewhere). Then stones are removed if they are in the opponent’s territory, and that’s it. There is no need to define alive, seki, dame etc.

2 Likes
137

All this discussion of the difficulties in determining disputed life-and-death issues in Japanese rules makes me wonder whether there are any servers that implement Lasker-Maas rules.

This is a ruleset that has scoring by territory + prisoners, not area, just like Japanese, and unlike AGA doesn’t have pass stones nor require that white pass last. However, it does allow playing out these life-and-death issues unproblematically.

Lasker-Maas as originally proposed has a superko rule, which is not at all traditional for a territory+prisoners scoring ruleset, but there is a version proposed that uses a long-cycle rule instead:

2 Likes
138

That retains the

can gain an extra point by not filling the last

consequence of Lasker-Maas, and the obvious way of removing that would allow either
player to use a double-ko seki to make the game a draw unless the other gives up the seki.
​
​
​My suggestion for removing most of that - though with infinitely many ko threats,
one could still get an extra point that way if the opponent doesn’t instead allow a draw - is,

the long-cycle rule may cause pass-pass to end the game even when there was a basic ko ban
just before it, but otherwise, pass-pass when there was a basic ko ban does not end the phase
and
pass ​ , ​ pass ​ , ​ capture that was basic-ko-banned just before those two passes
gets used as a fallback criterion for cycles in which the players pass the same number of times

.

1 Like
139

I wrote that variation article :slight_smile:

Since then I found a very interesting version that would be the basis of my preferred ruleset if I was emperor of international go and could simply dictate a ruleset for everyone :stuck_out_tongue:

Play lasker maas but the second phase (placing prisoners instead of new stones) starts IMMEDIATELY after the first pass. If we interpret the first player pass as picking the button (thus the first pass and grabbing the button should have identical consequences for ko-bans in the rules) this can be mathematically proved to always yield exactly the same score as Button Go up to a fixed constant factor (so setting Komi appropriately would give identical result in both versions always). As button go uses an area-scoring method with a slight adjustment due to the button, this is a very interesting way to get a ruleset with 100% equivalence between an area-counting method (button go) and a territory-counting method (almost-lasker-maas), like AGA rules. But while AGA rules almost always gives the area-scoring result, this modification almost always gives the territory-scoring result instead (but without potential anomalies like pass-fights not present already in area scoring, as it is equivalent to button go).

The idea of the proof is: taking positive score to be good for black (so score is black score minus white score). If black passes first, then we get “areascore + 0.5” in button go and would get “areascore” in this modified lasker-maas. If white passes first instead, we get “areascore - 0.5” in button go, but we would get “areascore - 1” in the modified lasker-maas (because the black move placing a stone right before the first white pass did not gave an extra area point!). So the result is consistently always 0.5 better for white when counting lasker-maas-like, than when counting button-go-like. Setting komi to 7 if counting button-go-like and to 6.5 if counting with this first-pass-lasker-maas modification, we are guaranteed identical results counting either way, always, even on pathological positions, and thus we have a territory-scoring equivalent of button-go.

1 Like
140

My preferred way to handle this shares some spirit but is a bit different (again, if I were rules emperor :P): The rule would be that first, any pass lift ko bans. Then, any time there are two passes it “stops the game”, but resumptions are allowed, so actually if any player says “hey we should resume”, the game resumes. So two passes only end the game if both players want to end the game. This can happen as many times as necessary.

However, if players enter an endless cycle which spans game-stops and resumptions like this, then when we consider ALL the stop-positions in the cycle and score all of them as final. If the same player wins in all of them, that player is deemed the winner. If however the winning player is different for different stop-positions in the cycle, then the game ends without result.

This rule is a bit uglier but it guarantees that there is no Zugzwang at all. For example it seems to handle in a more “traditional” way pathological cases like this, ending in “no result” with correct play when the result of the game depends on such strange ko, instead of degenerating into a no-pass-go game.

3 Likes
141

score all of them as final.

using Tromp-Taylor? ​ Tromp-Taylor Rules at Sensei's Library
If this would instead use hypothetical play or an encore, then
what about when it happens in hypothetical play or an encore?
​
​

guarantees that there is no Zugzwang

Note that this is via “endless”. ​ ​ ​ If ​ scoring “ALL the stop-positions in the cycle” as final ​ instead
triggered after a finite number of moves, then there would almost-certainly still be zugzwang.

(You may point out that there are only finitely many positions, so the players could in
principle give finite lookup tables for their strategies, from which it could be determined
whether there would be such an endless cycle. ​ However, in that setting, my suggestion
could also be limited to endless cycles, in which case it too would have no zugzwang.)