Would an auto chat entry do the job?
“Click the stones/territory to mark them black/white/neutral.”
And in non-bot games “Or restart to clarify position by playing it out.”
7 Likes
Fox only works for Windows and mobiles.
Let’s stay put on topic, how fox scoring is better as OGS scoring ?
The OGS scoring of the position in the diagram is not “terrible”; the game isn’t finished, that’s all. If it’s Black’s move, Black wins easily by playing in the bottom right. If it’s White’s move, White can probably win by a suitable defensive move in that area.
You can’t expect an accurate score of an unfinished game.
OGS does have some bad bugs, but this is not one of them.
The game was technically finished, because both players passed. The only thing left to do was scoring the final position.
The final position implies that both players considered the lower right to be white territory. Its boundary is closed, so there was no reason to overrule that.
12 Likes
I would rather say that the game is over because the boundaries have all been closed; the players have passed because the game is over, there being nothing more to do.
The act of passing twice doesn’t end the game on its own merit.
The problem is that Black does not know that the game is ‘unfinished’, since passing indicates that they don’t believe the area can be invaded. The scoring AI then shows that it thinks the area is not unconditionally alive, which may prompt Black to resume the game and actually invade.
The AI is basically leaning over the shoulder of the players and saying “I’d protect that area if I were white…” Which is without a doubt a bad thing.
8 Likes
Article 9 of the Japanese rules state that the game stops after consecutive passes.
A player could then ask to resume the game. But if that doesn’t happen, the position is scored as is (after agreeing on dead stones).
Players are not required to close their territories before passing. If they don’t, their unclosed area is just neutral.
9 Likes
We also need to be careful about the difference between rules and convention I think. By convention we don’t fill dame when playing Japanese rules but the rules state that leaving them unfilled makes the adjacent stones seki and the points they surround not territory.
There is surely then a similarity with leaving some minor borders unclosed and just agreeing to score it anyway (although this is not the convention).
In this case each player has passed implying that they are both satisfied that passing is the best move they could make (any other moves would cost them points) and so there is clear agreement on the state of the lower right and the rules are clear as to how this should be scored (points surrounded only by alive stones of one colour are territory for that colour).
Any other outcome is a system error it seems to me. Due in this case to the scoring system not recognising the difference between a game in progress and a game where play had ended (i.e. two consecutive passes have occurred)
7 Likes
Not to mention, as stronger players we might think that closing the borders is evidently necessary and that weaker players should learn their lesson; but the fact stays that there are a lot of weaker players reporting such games with open borders and finding it unfair that their game-long effort was denied because they didn’t spot a single open border. If both players agree that it’s unfair, what am I to do as moderator, saying “sorry to ruin your fun, but you’ve got to git gud”?
So far I’ve had no complaints from any player yet about their “win” being scrapped after their opponent forgot to close the borders and missed out on a large chunk of territory. To fall back on a rule technicality defeats the purpose of having a fun game. This isn’t a tournament after all.
If I were playing a friendly game on a real board with someone, and we discover a hole after both having passed, we would just plug that hole there and then and continue with scoring.
4 Likes
Doesn’t it show the unclosed area as neutral before accepting the score? So they are not paying attention twice?
1 Like
The interface doesn’t allow for the territory to be marked, thus we get a lot of cases where the reporter believes the system doesn’t function correctly (instead of finding the missing stone). Then there’s cases where the game is scored automatically (there’s an auto-score button, or against bot opponents).
It’s not really “git gud” though is it? It’s “get as good as a noob Uberdude taught the rules in person for an hour or two”. Here’s how I think it should work:
- During scoring confirmation open boundaries are clearly shown as making areas neutral not territory (lightvector has a nice UI improvement suggestion for this of animating from the hole)
- Players should resume play to close borders
- If they don’t and accept end position with unclosed borders, then those areas are neutral. Those are the rules of scoring. Do not change this.
- If they complain to mod, the mod should not annul the game, but teach them about closing borders, either via a demo or by directing to high quality tutorials.
The outcome of this process is education and beginners learning how to score, not preserving and accommodating ignorance. These problems seem more common on OGS than real life or KGS, which hints that it is the bad scoring implementation here that contributes to it.
6 Likes
It’s just the way online discussion works. From another thread:
1 Like
Something like this would be nice in the scoring phase:
Mark those empty intersections next to both Black and White stones in a very obvious way.
16 Likes
But then we go again into the limits of solutions given by AI. And it’s not always trivial like this(life and death, assumed weakness …)
2 Likes
You have to wonder why the OP posted a query about the behaviour of OGS on Reddit when this forum exists.
3 Likes
I don’t think you need or even want an AI for that kind of feedback to the user. A much simpler algorithm should be used.
3 Likes
OK, so, something quite big that was not said is that improving the current status quo is quite costy and dangerous: it involves developpement, and might do the wrong thing in the corner.
I propose something that should be simple to implement: use the AI score estimator 2 times by changing the playing turn. If the result is the same in the 2 cases, the score is OK. Else, these is a problem.
Some possible proposition for the problematic cases:
- just log them, to let some analysis after a month to have a better idea of what to do in these cases
- refuse to automatic score the situation, forcing the players to resume in case of a bot game
- warning, score estimation is confused, maybe your game is not finished. You can resume the game, call a moderator or force automatic scoring.
- notify a moderator to check this game.
- something else?
1 Like
This is basically what I’m suggesting:
As for logging them, see the post after the one I quoted above for some examples as well.
Forcing the players to resume or warning the players that the game is unsettled is similar to the current problem, in that it gives the player a hint that the board isn’t settled. You don’t want to give players hints that the board is unsettled if they don’t see how to invade, since that is essentially outside assistance.
Moreover, forcing players to resume will without a doubt result in many players just passing again, and a lot more of giving up and timing out when they can’t get the game to finish. Finally, forcing the game to resume is in no way in alignment with the rules of Go: the game should finish after two passes.
Notifying a moderator for each of such games is practically problematic, since we don’t have the manpower to deal with each such situation.
6 Likes