Sounds like an excellent explanation. I was unaware of the event you mention, as I haven’t been online much the last few days.
1 Like
How is it different from the normal behavior in which it attempts to finish the game with AI?
The current autoscore has done well in not scoring territory with open borders. However, it still has major issues by scoring based on hypothetical tesuji rather than by the board as it stands, which has been much discussed with examples in this thread and in the earlier thread that I started shortly after the 2021 change.
1 Like
Examples like this with no game link are not very helpful. The player(s) have almost certainly made manual changes to scoring anyway but how to know?
2 Likes
Sorry, I find it inconceivable that any players, let alone comparatively new players, would score the board in a way that mimics the old autoscore algorithm. In particular, the dame on C1 would indicate a recognition of the cut on D2. If either of them recognized that, they would have made the cut or defended the cut, respectively, not mark C1 as dame. Also, notice the puzzlement by both players in the chat about the weird scoring. This is a bizarre autoscore mistake.
3 Likes
I think this would be a proper marking of the final position:
I left out blue markings, because I don’t really know the difference between unmarked intersections and blue marked intersections. Aren’t both just dame points?
One could argue about scoring M13 as a point for white or not.
If I’m not mistaken, this would be a white win by 6.5 or 7.5 points (depending on M13).
2 Likes
Yes but why did autoscore at the “end” of a game come up with a score estimator type outcome?
Thankyou for the link Uberdude. I can’t explain the scoring but I can at least confirm that scoring was not modified by the players.
2 Likes
The current autoscore typically leaves territory unmarked when its borders are not closed; it does not mark it as dame. Trying to interpret the outcome like this case is what the previous autoscore system did.
Just for a bit of balance, it’s not only OGS…
9 Likes
6 Likes
This happens all the time under the current OGS scoring. As I have said many tines before, you must not only beat your opponent but you must beat the scoring AI, even when the opponent doesn’t see it.
2 Likes
I agree, but this thread serves both as an archive of numerous examples, and as a periodic reminder that this remains an issue.
5 Likes
A section buried in a long wiki page on Github hardly qualifies as expected reading for all OGS players. I’ve been playing on OGS for years and had no clue that existed. If players are expected to read that, it should be linked prominently from within OGS’s web interface.
3 Likes
There are many previous examples in this thread and in the thread on the same subject that I started before this one.
Scoring breaking the rules of go.
https://www.reddit.com/r/baduk/comments/14esk0w/a_recent_game_i_had_playing_black_what_are_the/
4 Likes