To be fair, there is a logic to the blue square behavior:
True dame is often just ignored, left blank as just empty space inbetween groups of different color. The Blue dame was meant to indicate that someone marked it as dame on purpose (though that “someone” could have been the system itself).
When I’ve had eyes in seki, or just a hanging ko, I’d appreciated that there is something to mark “Hey, this is dame, not territory”. It is somewhat unclear what happens to points that are just blank (and sometimes you don’t even know why it is blank).
In short, I think it is meant to be de difference between default behavior, and user input (or system input for that matter).
That said, this behavior is not consistent, so everything seems very buggy and confusing. Specially for beginners.
The graphic show the score estimator, not the autoscore, I believe. The blue squares are predicted dame based on whatever obscure playouts the AI did, while the unmarked points are unsettled, based on the same playouts, because the AI can’t predict any further. That graphic is not an example of the terrible autoscore.
Edit: I retract my statement that this is the score estimator. I overlooked the fact that the stones are unmarked. See below for my revised opinion.
Thanks for the added information. I take back what I said in my previous post. I overlooked the fact that the stones are unmarked, so it can’t be the score estimator, which uses Chinese scoring. However, it does not score the way the autoscore does today. So the possibilities are (1) that it is an old game, going back to the time before the autoscore update last year (the old autoscore did score like this, although this example looks much worse than anything I remember seeing); or (2) the game wasn’t even on OGS (although I see that the poster claims it was). I see no date for the post on Reddit. Do we even know whether that post is recent?
B+2 in that corner is also what would result from correct play under strict Japanese rules,
since Black capturing the 2 stones will force White to play another in Black’s eye.
Are you arguing that it is correct for the system to give extra territory points to a player that they shouldn’t have, in order to compensate for the prisoner points that they did not get because they didn’t play the most optimal sequence?
Also, I came upon this position specifically because I was trying to explain to a player in a review, that since this game was played under Japanese rules, they should add a move at O1 to force White to sacrifice an extra stone.
But it’s hard to explain why they should do that, if the system actually gives territory points in the seki as if the game was played under Chinese rules.
I think this is exactly the sort of situation where the players (or at least one of them) need to be aware of the real status of the group and designate the appropriate points as dame.
No; I’m pointing out that these are points Black should have.
In this case, the natural scoring is also a shortcut:
If White draws things out by not accepting it, then Black resumes and captures.
I do realize the only way that helps with the issue you mentioned is,
“The players went with this to speed things up, because it makes no difference here.” .
I imagine the walkie-talkie eight sekis are the simplest examples in which https://senseis.xmp.net/?WalkieTalkieEightInTheCorner
possible resumption won’t make Japanese rules give the natural score.
This game was played with Japanese rules. With Japanese rules, there should be no territory points in a seki.
Here, the system incorrectly gave 3 territory points to Black and 2 territory points to White, in the seki.
I agree that if the players had played differently, then Black might have gained one extra point of prisoners by forcing White to sacrifice one extra stone.
But:
The players did not play that way;
Even if the players had played that way, they shouldn’t have had any territory points in the seki.
The system scored that game wrong. It gave territory points inside a seki in Japanese rules, which is wrong.
No, they didn’t. Black did not capture the two White stones during the game because they didn’t realise that doing so would result in one extra point for them. And the players did not contest the scoring of the game by the system because either they did not notice at all, or they trusted the system’s scoring more than they trusted their own knowledge in the rules.
Which is understandable. If you’re playing a game on an internet server, it would take some arrogance to say “clearly the system is wrong here, I know better than the system” unless you have some serious experience with the game outside this server.
Here what was given is black +5 and white +3 (including the stones incorrectly showing as captured), which functionally arrives at the same score as if black played the capture and the seki was correctly marked. It is possible the players accepted this as a shortcut to the real score.
On the broader point, I think attempting to make the system definitively address edge cases creates more problems than letting the players mark dead groups and dame.
This is not an edge case. It is a simple seki. And the scoring system apparently correctly detected that it was a seki, since the dame is marked with a blue square.
There is no “shortcut” here. It’s a straightforward bug. Seki in Japanese rules don’t have territory. It’s as simple as that. I don’t know why you’re trying to argue so strongly that the scoring system was somehow right to be wrong about this.
I was not able to find anything addressing this more directly than https://senseis.xmp.net/?SekiWithEyesQuestion4
, but my understanding is that in the same way, there should be no dead-stones points in a seki.
The system also gave 2 dead-stones points to Black and 1 dead-stone point to White, in the seki.
regarding 2. in your earlier reply to me: Does OGS have some way to mark a
stone as dead without marking its intersection as territory for the other player?
regarding your more-recent comment:
We’re trying to argue that so strongly because this thread is full of examples
of pass-pass when at least one of the players shouldn’t have passed, and
discussion of how OGS’s scoring system should handle such situations.
Do we know that for sure? Or did one, or the other, or both, players manually “correct” the scoring?
The autoscore used to regularly give points in seki under Japanese rules. That was one of the problems that was fixed in the update last year. I can’t recall seeing, in the past year, a case like this where it made a mistake on that. It may be a one-off for some reason, but this is not the usual functioning of the autoscore today.
Wow! This is an extraordinary screw-up. This is how it used to score before the 2021 change. Either the autoscore has been deliberately reverted to the pre-2021 algorithm (which seems very unlikely) or this is a major bug. But why should such a bug occur suddenly?
Could it be related to what happened the other day? We had site-wide issues, and from how quickly they were solved, it felt like @anoek might have resorted to activating a back-up server or an older version of the code, just a hypothesis though
