Agreed, ideally. I’ll have to ponder on that a bit.
1 Like
I tested hexahedron’s seki test position on the beta server and the autoscore algorithm seems to adjudicate all group statuses correctly, and consequently the seki detection also properly detects dame properly, including teire (with the caveat that it scores points for dead stones left in seki eyes, as by commonly used server conventions)
… to be compared with hexahedron’s (from manually marked dead stones) …
Edit: on a 2nd look I do spot a teire around D9 which is not marked dame in both cases, although it does mark the teire at L8.
But it seems this is intentional:
3 Likes
Two things may worth noting:
1: Omitting dame play under Japanese rules is legal and an officially endorsed practice.
2: There are good reasons why the rules say no territory and no stone removal in sekis. Counting territory in sekis would of course be broken, but even w/o counting territory, just removing non-alive stones from there is NOT safe theoretically (it could change the cost of some moves into some rare sekis).
Of course servers often simplify things and don’t worry about rare cases (IGS doesn’t even differentiate sekis), but it seems best to lead users to the correct direction - which region is territory (closed off and fully/independently alive) and which isn’t.
2 Likes
I’m afraid I’ve not managed to follow everything but I do prefer the option to fill the whole area with the red squares because:
- It’s obvious something is not right
- It doesn’t point to a particular place and hence isn’t so much like outside help
- It doesn’t point to a particular place and make us/the system look silly when that isn’t the right place to point
I find it odd that we would point to one particular spot and say “play there” when for everything else we are keen to avoid giving hints to players.
And if the place we point to isn’t the “right” one then there might be some other unintended consequences that we’ve not foreseen, beyond just looking silly.
5 Likes
No, it’s not. This has been explained to you in detail fairly recently.
Not filling in Dame has implications for the scoring. Therefore the Japanese rules demand Dame filling before the scoring can commence.
1 Like
I have a strong disagreement with this, though you’re not alone thinking that way.
In my view there is a fundamental difference between open borders and life and death situations.
Open borders is only a beginner issue, and usually stems from a misunderstanding of the rules rather than a lack of skill. The priority is to help beginners understand what’s happening and help them get past that hurdle as fast as possible so that their games may be scored « as intended ».
By contrast, misreading life and death can happen at any level and is entirely a matter of Go skill. The focus here should be on making sure that we are not interfering with the player’s skill.
Filling the whole space with red squares seems the worst option to me:
-
if we agree the point is pedagogical in nature, it’s much less efficient, and more confusing than simply helpfully pointing the open border to the beginner. Especially in a situation where there is a single open point somewhere that causes half the board to be red…
-
and if you disagree and believe the focus should be on being impartial (like for life and death), then we should not show a red square at all and simply score the game as is (with large dame if territory is not sealed). That’s fine for me too as a fallback.
8 Likes
I’m not sure I disagree with your main point but I can certainly agree with this option
3 Likes
I can understand wanting to help beginners and distinguishing this from helping non-beginners somehow. I’m not sure how to achieve this in practice. I don’t think it’s so unlikely that a couple of non-beginners might miss a boundary point. Or a missing boundary point might have some life and death implications.
So I guess my assumption would be that we can’t ever really know that a person we are helping is a beginner and so I feel it’s safer to not help.
I suppose I would favour a system of ending the game and dealing with queries after.
It would be a whole different thing but I wouldn’t mind scoring all games automatically and on the finished game screen having a thing like “not what you expected? Click here for info” that would take you to this AI marked up board with explanations for the various symbols. And from there “need more help? Ask on the forums or report this game” with links as appropriate.
A final thought is that is were we to formally say that beginners getting help is ok then that makes it hard to object to beginners (or anyone!) saying “I’m a beginner and was using AI to help me learn where to play so I can figure out the rules” or some similar argument!
Not filling dame is not the same as not filling dame in the game. The latter is common, and the rules explicitly clarify that this is ok (without affecting seki detection).
J89 also mentions agreement about L/D and territories (thus sekiness as well). In an online game (w/o dispute) the stopped position remains untouched and is scored by the server, without manual scoring steps like backfilling prisoners, informal damefill or rearranging territories.
2 Likes
I can understand your view, but my point remains that either we help beginners on closing borders (at the risk of helping other players from time to time), and in that case we might as well do it properly; or we decide not to provide any help and we should get rid of that red square altogether (but it doesn’t seem to be the direction Anoek is going for).
I’m not sure I get this argument. I don’t think anybody suggested that “beginners getting help is ok” in a general way.
This wouldn’t change anything to the OGS rules, in particular on the use of bots.
2 Likes
No they do not. They explicitly state the exact opposite. Please just read the rules.
The game can only end through confirmation of L/D. The confirmation of L/D requires that the Dame is filled in. So filling in Dame happens before the game ends. There is no ambiguity regarding this.
Backfilling prisoners or rearranging territories are indeed scoring steps that happen after the end of the game. They only serve the purpose of calculating the score.
But as I’ve told you several times by now and as the rules clearly specify, the “informal damefill” happens before the end of the game. This action has an actual impact on the score of the game.
2 Likes
It’s probably very marginal but it’s not unknown for people to claim “I wasn’t using AI to cheat, I was using AI to learn”.
But you’re right. And I suspect that when this does happen then usually a few words of advice is all it takes to avoid future issues.
Most games have no actual confirmation phase, agreement suffices. The rules say “ends by confirmation AND agreement”, in a bit loose phrasing (and not just on L/D but also on territory/sekiness), but if you read the whole text it is clear what confirmation refers to.
And with Asian rules actual practice matters as well. The commentary endorses game stop without dame fill since this is so common that written rules also had to incorporate it (in a flexible enough way that doesn’t conflict their seki logic, hence some vagueness). And in an undisputed server-scored online game NOTHING happens to the board position after the stop.
1 Like
I found out what was causing that, it should be fixed now.
3 Likes
I now make it so if there’s a snapback situation, we leave it alone and let the players sort it out, so the board for that snapback test now gets auto-scored like this:
10 Likes
This is looking really great; thanks for all your work on this, anoek
8 Likes
Thanks for checking!
3 Likes
It’s always great to see that there is continous development for this platform. The new scoring mechanism seems to be much better than the legacy one.
The idea to give hints for beginners to close borders is innovative and might finally give AI-assisted manual scoring (AIAMS) a use case.
There are still two things about the new version that I’d like to address.
1) Proposing “wrong” answers
As can be seen in the image below the players get an AIAMS proposal that seems off. In this kind of scenario Fox-style auto scoring would have just retrieved the “correct” result. While a gennan-style manual scoring would have left it open for the players to decide. The AIAMS however proposes a solution to the user that is “wrong”.
I think this could potential confuse beginners as well as veterans that are new to the site.
2) Giving hints
AIAMS giving hints is a widely known and criticized issue that still occurs with the new version. In the image below White gets the crucial information that he needs to connect at D6. He can continue with the game and use this new information to alter the outcome of the game.
1 Like
How would Fox-style auto scoring score this? Sensei’s says that " the Japanese ’89 rules make this seki if it is left on the board at the end of the game.", so this seems correct, no?
Re #2, that game is so far from complete. If you have two brand new players passing in this state, then they need all the help they can get. I don’t even see the marked dead stones as particularly problematic in such cases, if anything it’s just a learning moment for both players as they learn a bit about the game. Them passing and attempting to score at this phase just shows that they really don’t know how to score and when the game is over. The system highlighting that they should continue playing (even if that means showing them a little hint about why they might need to keep playing) is a good thing I think. My hope is that after going through this a time or two we’ll quickly bring them to the point where they have a better understanding about sealing off territory and how to play the game, and when that happens they’re not going to be experiencing the sealing system that much.
9 Likes