A compendium of OGS's terrible scoring system confusing beginners

Feedback
81

I think the issues raised by this thread are important, but the discussion has become quite sprawling, since there are actually several distinct (although related) issues regarding scoring. Here’s a non-exhaustive list of various concerns posed in question formats:

  1. Does the scoring system provide enough assistance toward automatically marking territory, dame, seki, and dead stones?
  2. Does this assistance offer any unfair advantage to the players in terms of recognizing unsettled positions?
  3. How should games with unsettled positions and unclosed borderlines be handled?
  4. Should beginners be given the “both players lose” when called for under formal application of the Japanese rules, or should some softer approach be used?
  5. Is the interface friendly and understandable enough to make it easy for players to understand how to mark territory, dame points, and dead stones?
  6. When playing under Japanese rules, should players be required to fill in dame (as they are compelled to under formal Japanese rules)?
  7. How should scoring be handled for games against bots?
  8. Should scoring be entirely automated by a strong AI engine like KataGo?
  9. How should scoring disputes be handled under Japanese rules?
  10. Should some sort of “play-it-out” procedure by adopted to approximate the life/death results of the Japanese rules?
  11. How should scoring mistakes and various forms of deliberate mis-scoring be handled?
  12. Should players have more tools to indicate that they’re protesting the score (for later review by a moderator), while accepting it for the sake of moving on?
  13. When should games be annulled under various forms of scoring mistakes and deliberate mis-scoring?
  14. Is convincing an opponent the mark your dead group as living (since they do not see how to kill it) a form of score cheating?
  15. Does the previous question depend on the rules set being used?

I haven’t read much of the rest of thread, and I’m sure the discussion has already touched upon many of these issues already. However, my point isn’t just to raise more questions or rehash existing ones, but rather just acknowledge that the concerns about the scoring system is not just one or a few issues, but instead a complex web of issues and questions.

11 Likes
82

Thank you for your summary @yebellz

Some easy answers:

When both players pass they are effectively declaring that all positions are settled as far as they can tell. If they later realize they were wrong in a game deciding way, both players lose. If they were wrong but it makes no difference, then it makes no difference (as the winner won’t mind restarting). If they don’t notice and go on to confirm the result, then that result stands as is, regardless of what KataGo or anyone else has to say about it.

In accordance with the rules, i.e. not territory.

Yes. And not just beginners - all players. However, I expect this is unlikely to come up much, especially for beginners, as both players have to realize the mistake and it has to be decisive.

Yes

No

In accordance with the rules, i.e. agree life and death, any player can request restart to confirm life and death, opponent plays first, players must pass before recapturing ko, once life and death is confirmed they score the original position.

If it’s an honest mistake, then no. If it is deliberate deception, then yes. If in doubt, play it out.

Only in so far as the life and death status depends on the ruleset, e.g. bent 4 in the corner with unlimited ko threats elsewhere Chinese vs Japanese rules.

1 Like
83

Agreed on all accounts with @dragon-devourer, except :

  1. When playing under Japanese rules, should players be required to fill in dame (as they are compelled to under formal Japanese rules)?

How would this work ? If both players pass without filing the dame, do they both lose ? This seems awfully weird, but I don’t see how they would be compelled to anything otherwise.

  1. Is convincing an opponent the mark your dead group as living (since they do not see how to kill it) a form of score cheating?

No. If they do not see how to kill it then for all intent and purposes it is alive (whether or not you would know how to kill if the situation was reversed).

3 Likes
84

Nope. But groups touching dame are alive in seki so “territory” they surround does not count as points which fundamentally changes the count.

Yeah, fair point. I’d go with that :grin: :+1:

1 Like
85

Let me address the weirdness of situations like this:

This would appear to be a situation where someone (possibly mistakenly, but likely deliberately) misused the scoring tool to mark several points inside White’s territory as “dame” points (as indicated by blue squares), which significantly reduces White’s score and possibly changes the outcome of the game.

One easy solution (as suggested by @stone.defender) is to just completely disallow dame to be marked inside of apparent territory. However, this would negate what I believe is its primary function, that is, allowing players to mark teire points left due to unfilled dame.

To explain let me use some examples from another post, where I discussed some of the intricacies of dame filling.

Under formal application of the Japanese rules, player are compelled to fill in dame points to prevent technical sekis. Of course, in practice, with casual play and most games on this site, players will often leave some dame unfilled, even sometimes when those dame kind of matter.

A basic example

Here is a simple example that we often see in OGS games that have gone to scoring:

image

The dame at A has been left unfilled, which has allowed Black to avoid defending B. Formal application of the Japanese rules requires both players to play another move (White at A and Black at B), however, players on OGS often skip these additional two rules as a shortcut. Hence, to properly score the above situation, the blue square feature should used to mark both A and B as neutral.

Some other examples

Teire points that should be marked as neutral can become more complex than that.

In the below two examples, there are no more urgent moves left to play, and conceivably the players might choose to pass early before filling in the remaining dame or fixing the teire. In both cases, the white stones marked with X are dead. Let’s consider how one might try to score these positions, if both players simply decided to pass instead of playing out the dame filling and teire.

In this first case, the two points at 2 and 4 should be marked with the blue squares, since Black would have been compelled to fill those in response to White filling dame at 1 and 3.

image
image619×532 43.4 KB

The White stones marked with X are dead and the points underneath them should count as territory for Black. However, would the scoring system properly count those points as territory (given that they are only adjacent to points marked as neutral with a blue square)?

In the second case below, the point at 2 should be marked with a blue square, since Black would be eventually compelled to fill at 2.

image
image253×554 20.8 KB

The white stone marked with an X is again dead here, but since Black should eventually be compelled to fill that as well (in order to connect, in response to White filling the dame at 1 and 3), the point under that dead stone should not be counted as territory for Black. However, can the scoring system mark the White stone as dead, but also the space underneath as non-territory for Black?

Hence, if the players take the shortcut of not playing out the dame filling in the last two cases, I’m uncertain if it is even possible to score the games correctly. Of course, in all of the above cases, the simple way to avoid any issues is to simply play out the dame filling, which would obviate the need to mark points with blue squares and worry about which points are counted as territory.

4 Likes
86

that someone was Kata Bot himself

I talked about not letting Kata Scoring to mark sealed territory as neutral.
Players should be able to paint as they wish.

87

I think players should not be forced to fill all dame under Japanese rules, but it should at least be recommended in the description of the rules. If you don’t fill all dame, some teire point could wrongly count as territory. Players need to be aware of that to properly finish and score.
But the scoring tool should not try to fix such mistakes. If players make mistakes while finishing and scoring, it’s their responsibility, just as mistakes made while playing.

5 Likes
88

Exhibit C: a 2k (so not a beginner!) who thinks you don’t need to close areas for them to become territory, because OGS used to let you be lazy and not bother and then magically score some of it for you:

2 Likes
89

Yup, I told you, this happens a lot, and generally people don’t think it’s their mistake.

This player seems to be quite unfriendly though. I’ll explain to them that their opponent cannot close their territory for them. In a way the black player could’ve made the problem more clear by playing J9, H9, G9, etc.

1 Like
90

It happens a lot on OGS it seems, presumably because the previous magical presumptive scoring system mislead them into thinking that’s how scoring worked? And/or there’s lots of people here who have learnt the rules on their own from some webpage or video and not had the benefit of an interactive lesson with a human teacher? I’ve not encountered it in other Go communities (real life in the UK, KGS) beyond total beginners, who you explain it to on a demo board and they quickly learn and understand. Of course they might miss closing a hole afterwards, but the response is always “Oh yes, silly me, I didn’t see that, I’ll close it now” not “I don’t need to close it, and how dare you suggest I should, argue argue hrmph hrmph”.

2 Likes
91

So, I guess we should blame the previous scoring system, even though at the time nobody seemed to have problems with the way it dealt with open borders. There’s not really a solution here, then…

92

Indeed, a time machine is the best solution. Unfortunately it’s a lot harder to correct mistaken thinking than teach it correctly in the first place.

2 Likes
93

I don’t understand how white could have missed that.
I mean, even if you miss the gap before passing, after passing, you get feedback that you have almost no points. So white should check if the score is according to their expectation. If it isn’t, look closer and resume the game when you notice your mistake.

94

but black missed the gargantuan seki (as well) :man_shrugging:

95

I see that white fixed it after all on move 49, but timed out and then the game was annulled?

BTW what are those triangles on move 47?

Edit: Oh, i see it was a leftover from the variation posted in the game chat.

96

Time (even sans time machine), will take care of this issue. The current autoscore appears to handle open borders correctly, and OGS seems to have a high turnover rate (based on my observations of the chat over a 2-year period). Therefore, the autoscore will, before long, inculcate most new players and veteran players into closing their borders (or suffer the consequences), even if they never learned the game in a formal manner.

1 Like
97

Yep, without someone explaining it to you, it’s really confusing. Go is easy to learn my butthole. You have no idea what’s going on in scoring even if you know you can change status of the groups. And watching/reading a tutorial maybe wouldn’t help because your understanding is so rudimentary. But at least you would know how the final position is supposed to look like.

I suspect for best experience one needs experienced player walking through a real game, like two beginners play each other and then game scored by someone else. But we can’t have that.

I like how it’s done on pandanet somewhat because it marks possibly dead stones to bring to your attention. On places like KGS you can miss sometimes one stone.

But OGS automatic score is really helpful to beginners too. Automatic marking allows beginners to not be totally confused right away and score the game somehow until they improved their understanding to score on their own. We focus on few side-effects here, but I imagine in most cases it turns out fine, beginners play, get smarter and can score the games on their own after a while. I think that’s how I learned though it was a long time ago. Would I be here without friendly auto score algorithm helping me?

I suspect they both know you technically need to close territories. Considering black, their opponent, passed first, I imagine they tried to perform a pro gamer move: they pass, opponent passes and they claim they win because opponent’s territory doesn’t count. Because they haven’t tried to close borders by themselves too and just passed multiple times. It’s stupid.

On OGS there’s a annoying habit of some people to “fuzzy score” the games. When they see they’re losing they pass instead of resigning and if you pass too you score the game as it is. The winner is correct but instead of resign you have a “score” and for some reason it’s better? Most of them resign if you refuse to pass.

I met a 3k (so 1k in new ranks) who didn’t understand П-shaped seki or something like that.

We can have the best of the two worlds. Auto score determines dead stones and simple algorithm fills the territories with color. That will teach beginners how dead stones look like, and that you need to close territories. There’re cases where it’s gonna be broken, but still pretty good.

Edit: or do we already have something like this? What’s the current system?

I’m generally for autoscore.

@Uberdude what do you think about Foxy scoring?

4 Likes
98

Here another example of scoring problem

Screenshot_2022-02-23-22-59-30-77
Screenshot_2022-02-23-22-59-30-77720×1520 175 KB

Both players agree but then scoring don’t. Andsubstract some of white points.

AI punish white even more

Screenshot_2022-02-23-22-57-40-60
Screenshot_2022-02-23-22-57-40-60720×1520 174 KB

5 Likes
99

I never saw people against autoscore, who wants to manually count the score in an online game ? :sweat_smile:

The debate is rather where to put the cursor between a strict application of the rules (area not closed is not a territory; stones on the board are alive unless both players agree they are dead), and a “smart” autoscore which tries to guess the intention of the players (white probably intended to close-off this area by playing here; the players probably believe those stones are dead since Katago see it as dead…).

The smart one can save time, but it has the downside of inventing moves that the players may not have seen (specifically for life & death situations) or supporting a wrong understanding of the rules (“you don’t need to close areas for it to count as territory”).

3 Likes
100

IMG_20220225_220507
IMG_20220225_220507718×396 130 KB

Another mistake (Japanese scoring)

1 Like