The big group stays alive and then you have some dead or unsettled stones out at the ends; the initial estimate depends on which random direction the AI plays…
But whatever happens, the dead ones become “dead stones in dame” and get unsettled, and all the unsettled stones live because it’s dame, so everything stays alive.
I don’t think you could ask for a better result in this case.
Really surprising to me where the AI puts the unsettled stones. Is it giving them up in a ko fight? And it only shows it that way about half the time. This is the first case I’ve seen where I can’t relate these pictures in any way to the OGS autoscore result.
In any case this new autoscore algorithm happily preserves the integrity of that group in the lower-left and would score everything as the players surely expected.
It’s weird that the current algorithm calls the group dead but not territory.
Anyway, the v3 algorithm would call everything alive in seki:
(Also it ignores the extra point black could have won in the center and snapback shenanigans in the lower-right.)
Is there a better way to handle it?
I find this picture a bit hard to read.
It feels like there should be a distinction between territory and dame, at least for readability, rather than lumping both together with a plus sign?
Lots of things to distinguish between:
And if you want to get technical about some oddities with Japanese rules, you could also distinguish dead stones that are not removed (for scoring purposes), and also consider if one wants to have different markings to highlight the eyes of seki positions (as non-territory) or territory points that must be inevitably removed (filled to defend) once external dame are eventually filled (i.e., in situations where players take the shortcut of not filling all dame, but want to score pretending that they have and made the requisite defensive moves). We need to go back to using a variety of fruit and vegetable emoji to help distinguish.
I’d like to see more fruits in there too, but all the algorithm does now is mark its decisions about dead and alive stones. I’m assuming that this would feed into some other thing that then counts the score according to the rules of go.
One major difficulty would be distinguishing seki from simple dame for the purposes of Japanese scoring. I doubt that it’s possible.
