The idea is to take a look at go rules from the view of expecting (reasonably) coherent gameplay and behavior between area and territory scoring. Basically, to find out what rules work - and what don’t and why - universally, regardless of scoring method. This simple requirement already exludes a few things so may be a useful guide.
Anybody with interest in rules is welcome to look at or edit the page. Is there something wrong, did I forget something, are there further rule components that could be evaluated on this basis?
For territory scoring but still counting points in sekis, what if we only count territory when no bordering stones are in atari? That should fix the mannenko situation, while still allowing counting of points in sekis for single point eyes or false eyes in all normal asymmetric-eyed seki positions.
The only difference would be cases where the final scoring position includes empty regions that genuinely “should” by Go players’ normal standards be counted as real territory despite the owning player’s border being in atari. Are there any such cases?
(Or, are there still positions that are clearly still broken by counting points in sekis with this alternative territory definition, in a way that they wouldn’t be broken by normal area scoring or normal seki-less territory scoring?)
The atari idea was recently suggested by @gennan (preceded by dead stone removal, to avoid problems with snapback or dead stones behind a ko). It seems to fix the common cases, but I’m not sure if it completely or theoretically excludes the possibility of a seki with different area<>territory behavior or stability.
For example, dead stone removal from sekis sounds risky (and unfair in anti-sekis), territory rules tend to be careful when they grant free removal. And area rules don’t make the atari exception. Maybe either of this could be used to construct a multi-state seki where the value of states and thus stability differs? But who knows, until someone shows such an example, maybe the atari rule actually is a complete solution / alternative.
Actually, it seems the atari rule would still lead to difference to both area and normal noseki-territory. Dead stone removal from sekis seems to imply a third type of behavior.
Under area scoring, ultimately all stones staying on the board are alive (and even ko stones make “territory” in sekis). Under territory scoring, seki stones involved in state cycling are usually dead (remain capturable in hypothetical play), but under noseki-territory this has no consequence (dead in seki = alive in seki, no removal and no terrirory either way).
However, under noatari-territory, these cyclic stones are removed from dame and then backfilled to territory elsewhere (or just added as points), so they have a half effect on score.
This normally still doesn’t make a behavior difference, because cycling stones tend to be symmetrical (like in a double ko seki), so the score shift tends to affect both players equally and cancel out. But consider a position where the switching stones are asymmetrical, like with a 0-sided ko (unfillable for both):
This is a genuine repetition under both area and territory. The two states are truly not of the same value under either scoring system. In area scoring taking the ko increases one’s area, in normal noseki-territory the ko itself doesn’t score but each capture in it gains a prisoner. Thus in close games the players need to continue capturing and recapturing forever.
However, under noatari-territory, removal from sekis seem to change things. A capture in the ko doesn’t gain a new prisoner anymore (the ko stone was dead and removable anyway), but even gives a new prisoner to the opponent (the new ko stone). So the situation is the opposite of normal area/territory scoring - capturing the ko actually decreases one’s score. So with a kind of reverse effect, noatari-territory turns a draw into a scorable loss.
Not sure if you’re describing my proposal here, but I didn’t mean to imply that a chain in atari would be removed before scoring (well it would if both players agree it is dead after passing).
I only meant that if the atari chain is considered alive by the players after passing, its liberty still wouldn’t count as territory.
This is an indirect consequence of the atari rule. Those stones are not removed because they are in atari, but because they are dead (remain capturable in hypothetical play).
In normal noseki-territory, dead stones are not removed from sekis (this exception is the essence of such rules), but noatari-territory has no special rule about sekis, so dead stones are naturally removed from there, aren’t they?
Suppose the players are good enough to understand the rules, so they have no disagreement, they mark dead stones correctly, and in accordance with the rules.
The mentioned cyclic stones in sekis are dead, this is the same regardless of the atari rule. For example, the ko stone above is dead, because it can be captured in hypothetical play (without enabling a new uncapturable stone).
The difference is only about how to handle dead stones, whether they are removed from sekis as well, or whether the rules contain the seki exception.
I think right there my idea differs from yours (I might be deviating more from Japanese rules than you?).
I’d say that the burden of proof is on the player who claims something is dead (not on the player who claims something is alive). And by dead, I mean permanently dead, i.e. even when ko bans are lifted.
So if black wants to claim that white’s stone at A4 is dead, black should be able to prove it with a sequence that is as close to normal play as possible (putting aside positions with unremovable ko threats for now).
To prove A4 dead, black would then need to capture at B4, white passes, black plays F3, white recaptures at A4, black passes, white connects at B4, black passes, white passes.
And then everything is alive (in seki), so black failed to prove that white A4 was dead.
So I’d say this final position should be scored as a seki with A4 being alive, but no white territory at B4.
Or am I missing a working sequence for black to prove A4 dead?
Yes, your sequence seemed to miss the fact that in this special position, neither player can connect the ko nor approach on the right without collapse. But this seems less important for the main point.
To prove the ko stone is dead, B just needs to prove he can capture it without negative consequences. He does not need to connect it afterwards or do anything more, since the essence of (currently used) L/D is uncapturable=alive, capturable=dead (normally, with provision for enabling, snapback etc).
If you mean to use a whole new different approach for L/D, you need to define it - but this would be a very tough nut with lots of pitfalls!
I assumed the atari rule would keep using the standard L/D. In any case, the two seems entirely different questions.
I suppose we agree that, at the very least, stones which cannot avoid being permanently removed from the board (regardless of who plays first) are dead.
To me, making a distinction between “removed at least once” vs “removed permanently” doesn’t seem like a “whole new approach” for L/D, although it does leave the question if stones that cannot be removed permanently should be marked dead when the game goes to scoring.
My feeling is that the answer to this question should be “no”, because it feels more natural to me.
It still doesn’t clear up cases such as you gave above, but I think your position should be subjected to the “long cycle” rule that applies, when neither player is prepared to go to scoring while the other has the ko advantage.
If the players agree to go to scoring with your position (which to me seems like a mistake by black), I think it would be fair to score it as a draw. How would you prefer to score it?
One reason current L/D is about “permanently alive” and not “permanently dead” is probably J89 L/D example 2. There two B stones in the corner are capturable, even permanently, yet not dead (as they give rise to new permanently alive stones nearby).
Sure, but this assumes that taking the ko is a step forward, or improves one’s score! Which feels natural and is true normally, both under area scoring and (noseki) territory scoring, so the players are interested in continuing the repetition.
But it is not true under noatari-territory scoring (at least with the standard L/D), where every ko capture decreases one’s score by one. I think this reflects some of the risks of not excluding sekis and allowing dead stone removal from there. Territory scoring is just safer without them.
I’m not against new L/D approaches btw (mentioned one here), but until you define a coherent system based on “permanently dead”, it seems too early to rely on it for analysis of a different idea.
To be clear: I’m in favour of scoring territory in seki (just as any other territory). Let’s call that “seki-scoring”. I don’t like to treat seki as a special case in scoring (no points in seki, which I think is what you mean by “noseki-scoring”?).
So my intention is using “permanently removable” in combination with seki-scoring.
Life-and-Death Example 2
This position is a seki. Black does not have to play A .
Reason why the two black stones are alive:
If White captures Black’s two stones, Black can play two new stones (at 2 and 4) which White cannot capture. Since there is a dame, the position is a seki.
Do you mean that the definition of “permanently removable” needs to account for cases like these, where white might claim that black’s corner is dead after passing, and the proof involves some trade that white didn’t want to make during the game?
I’d say that if white tries this under seki-scoring, black can just resume and throw in at A to prove that his corner stones are alive. To avoid such shenanigans during scoring, black can even throw in at A earlier in the game (that throw-in wouldn’t change the score under seki scoring and it’s even a useful ko threat).
But there are no stones in atari in the 1st diagram, so I don’t quite understand how that example relates to “noatari-scoring”. What do you mean by “noatari-scoring”? Are you referring to white’s atari at 2-2 in the 2nd diagram?
I’m not quite sure what you’re objecting to with using “permanantly removable”=“dead” (and the rest is “alive”).
Do you mean that you’d rather have the rules define “alive” than define “dead”? Like “permanently unremovable” = “alive” (and the rest is “dead”).
Ok, but this misses the point (B should be alive even without throw in). See also L/D example 4 for a similar case, where “permanently capturable” still doesn’t mean dead (and no throw-in is possible to cover up the issue).
If you mean J89 example 2 and 4, those were mentioned for your “permanently capturable” idea. And the atari rule itself never played a role in any of these examples, but the atari rule implies that dead stones are removed from sekis (since there is no seki exception). This is what causes problems.
The 0-sided ko example shows the risks - it leads to a behavior that is different from both area and normal territory scoring. Basically, if dead stones would be removed from sekis, then playing into a cyclic seki (flipping a ko for example) would lose a point, just like playing into opponent territory does. This is very different from common sense, area scoring (where all stones on the board are alive) and normal territory scoring (where it doesn’t hurt the score to have dead stones in seki).
Again, this assumes standard L/D (based on permanently alive), where cyclic ko stones are dead (as they remain capturable, can never become or raise uncapturable stones) - but I yet to see a working L/D definition based on “permanently dead” (eg. J89 example 4 mentioned above).