I guess I don’t really understand, or see the value of, the concept of a game being ready to be “properly scored” or not.
As a beginner I was further confused by this new, vaguely defined “practical concept”, until I just learned to dismiss it entirely.
If the goal is to avoid confusion, what matters is to have clear and consistent rules. When a beginner shows a game or a theoretical example, being answered “you can’t score that” is both untrue and unhelpful, it really makes the game sound more complicated than it is.
Now I understand that “avoiding confusion” is not the only goal OGS is pursuing. Another goal may be “be convenient and minimize players’ efforts in the scoring phase”, and on this ground I can understand the need for compromise.
This may be a bit tangential to the discussion above, but I think in the example above with the red square, black should get the point for the red square. And likewise the purple square should not be a problem for scoring.
In both cases, with sufficiently good players (well, anyone paying attention really) those points would of course be taken and if we were estimating the final result from an unfinished position then that would be a fine estimation to make.
But we’re not estimating the final result of the game, we are scoring a game that has already been finished, according to both players. The players have passed, indicating they don’t see any productive moves to make. They have failed to notice the atari so why should the auto-scorer tell them about it? And even more so for the protecting of the cut if the atari should be protected - there are game situations where you don’t have to protect at P18, if the player might have (mis)read they don’t need to protect or the opponent failed to notice the cut, then should the scorer tell them about it?
I’ve been working on an auto-scoring implementation and I’ve convinced myself that any unsettled stones/groups (that can be made alive starting from a move by the group’s color) should be considered alive and well. Only groups/stones that are dead from either players’ move should be marked dead.
I think showing the players where two potentially separate territories/areas meet without any stone in the way is OK. But I would prefer to leave the ataries/cuts on board and trust the players’ apparent judgement about them. Unless we do full AI auto-scoring.
There’s also the quirk of Japanese rules that requires dame filling, but I don’t think that this is just a Japanese rules feature. It sounds like OGS will have a way to report to the players that the game was not “properly scored” because of some open border.
How can you detect that programmatically? Maybe it will be something like, Katago thinks you can get some territory, while according to the actual rules of Go it counts as dame? If so, that implies that our scoring system would finally know the actual rules of Go, which would be great.
Note that if you score a game manually on OGS, you need to click on every dead chain manually, which can be a bit tedious. This might be one of the reasons why the autoscore is popular.
Compare with KGS, and other pre-AI go servers: there, if you clicked on a chain of stones to mark it dead, every other chain of the same colour inside the same contiguous territory was marked dead as well, which made manual scoring much faster and much less tedious.
Oh that’s interesting - the OGS in-game estimator actually does the “contiguous territory” thing (search toggleMetaGroupRemoval in the goban repo). Though the in-game estimator is actually where you don’t want this behavior because borders are usually open.
(this is about to change in that v0.8 update @GreenAsJade linked though)
Yeah I think that meant usually that even toggling one or two stones would toggle most of the board as dead, if those two stones were “dead” but not fully cut off or surrounded from everything else.
Not only does it make things faster (scoring on KGS takes literally 5 seconds, so all of this extra confusion and inconsistency and not following the rules of go on OGS to make it “faster” or “easier” doesn’t ring true), it emphasises the point that if you end a game before closing all the borders, then silly things happen (as in what you thought should be your territory, isn’t; it’s sensible given the board state) so you shouldn’t do that and you quickly learn about the need to close borders, rather than continuing under the delusion that you can finish games with open borders and get some AI to give half of a contiguous region to black and half to white with a few dame in the middle.
