Scoring counts in intersections as units, but one way or another all conventional rulesets ensure that neighboring empty intersections score homogenously (even LJRG, with its way of definitions).
I doubt itās circular with āpassiveā. If W were in B territory, B could capture without any backslash or negative consequence, and keep all territory. But it still seems better not to define life and death at all, only define territory.
As discussed on L19, in some cases the threat is not usable in actual play (like costy collapsing of a seki), while in other cases the threat is usable thus bent4 could not be captured in reality. So without area playout, a static territory view will always be incorrect in some cases.
Japanese-style āall threats invalidā leads to a scoring error of dozens of points in cases the threat was valid (by incorrect removal). Hypothetical play with normal ko - which means āall unremovable threats validā due to the enable rule - leads to a scoring error of several points if the threat was invalid (by forcing the capture to be played out in game).
But since a regular bent4 works well in hypothetical play with normal ko (including threat removal to avoid enabling on ignored threats), and since both errors above seem similarly bad, the simpler route seems tempting.
Leaving out sekis from territory actually keeps the two scoring methods closer. It makes TS robust and avoids larger behavior differences (like territorially unscorable positions). And control based territory view of sekis seems natural and convincing.
2 Likes
[Replying to the original post, disregarding all other posts that follow]
White wins because to come to that board position, black should at least pass two times.
Leaving out sekis from territory actually keeps the two scoring methods closer. It makes TS robust and avoids larger behavior differences (like territorially unscorable positions). And control based territory view of sekis seems natural and convincing.
If defined rigorously, as it can be (via ācontrolā) this is one part of TS I like. But I donāt see how to adapt it for AS without major complication (relative to the rest of AS rules).
I meant leaving out from territory without touching area. It leads to a point or two of score difference sometimes, but we already have that on Bās surplus stone and area point on last dame. In exchange it prevents more serious differences in behavior.
The way it is done traditionally do make sense imo.
Ok, Iāve found the reference to regions in the LJRG rules now and you might have a point about how these are defined. Robert Jasiek in his commentary to his new amateur Japanese rules however defines control by each intersection. See: ajcom
āthree-points-without-capturing adjacent to a seki shapeā for an example. I also think this is the only correct definition which would make LJRG problematic to apply in some cases.
Iām going to suggest that the game I posted is such a problem case. There is a way for Black to capture a two stone string from White. This string however is only one of the walls surrounding the top left territory with some of the other walls being alive independently. If you look at the control definition White can make two eyes for a group on this top left region, so one choice of resolution for this situation is to accept that this whole region is enclosed by White and it Whiteās territory even with the teire flaw in the wall. The score for that is the same as the score when neither player notices the flaw. Another way to resolve this would be for the control to be able to be defined as a sub-set of the territory, in which case White has control over most of the region as could be established. In this case the game score could be resolved to match the AI score, even if its scored with the teire unfixed.
My basic contention here is that the control set does not have to match the set of neighboring empty intersections in cases with these teire flaws so any rule sets which want to treat them as identical will have some issues scoring these cases.