I kind of get that sentiment but if two players play chess and don’t realize it’s checkmate any chess site there is will still end the game. Sometimes it’s better to just throw people into the cold water.
Never seen that. Do you have a reference or a screenshot?
This wording make it sound as if we’re throwing beginners into the proper way to play, and the only issue is it’s a bit too sudden. But relying on AI to determine the status of groups is not the proper way to play.
I see it more as another type of “playing with AI support”, to use the poll’s wording.
Still, I understand that it’s quick and convenient.
Never seen that. Do you have a reference or a screenshot?
Though I voted manual scoring, I would have no problem either with full autoscore personally. FlyOrDie started doing it some years ago and I’m quite happy with it.
I’m not overly concerned with novice players sometimes being confused by autoscore results, as long as those results can be easily explained by stronger players without invoking excuses like “Yeah, autoscore works in mysterious ways” or “Yeah, autoscore messed up with that one” (as seems to happen frequently with GoQuest/BadukPop autoscore questions on Reddit).
A full autoscore algorithm should be of very high quality for me to prefer it over manual scoring. But OGS doesn’t seem to have a great track record when it comes to prioritising to fix quirks and bugs in its implementation of the scoring process.
So my concern is that OGS might settle for an algorithm that is not really accurate enough to score higher level games correctly, deviating from how strong players would score a proper final position. Or deviating too much and too often from how stronger players would score an “unfinished” game of weaker players.
This is a temporary thing for beginners. It will resolve itself by players learning from the output. This is a negligible edge case. I’ve never seen anyone on Fox saying “The game was automatically scored correctly but as a beginner we wanted to score the game incorrectly. Please change this!”.
No. There has to be a very clear distinction here between those to approaches. When using auto scoring two humans are playing, then the games ends and then AI outputs the result. So the score calculation is AI based but the playing part is not AI supported.
The current OGS feature, on the other hand, provides the AI input during the game, before it is concluded.
Very important difference.
Ok, so the feature is there but the option is hidden until it’s actually available. Which makes sense. Sounds like a great feature!
tldr, I think it would be a great feature if the fixes I suggest in the linked posts were implemented. Care to weigh in on that thread in support for one or more of them?
Closing borders is a beginner issue. Misreading life and death is not, it can happen at any level.
“The game was automatically scored correctly but as a beginner we wanted to score the game incorrectly.
That’s a wrong way to present it. AI scoring is not “correct”, it’s the player’s view, their mutual agreement on the status of each group, that is the correct result as per the rules.
AI scoring is a shortcut, and in the vast majority of cases will match the player’s view, but if there is a difference then the “correct” result is the player’s view.
Very important difference.
In both cases, a player may win or lose thanks to the AI revealing something neither player saw during the game. The process is different, and I do prefer automatic scoring over current OGS system, but it’s still AI-interference.
Thanks for linking that. I will do that as soon as I have some time.
I agree it is an AI interference. But it only affects the interpretation of the result. It does not affect the actual stones on the board. Which I think is a very important difference.
I’m just wondering if manual counting is worth the effort but I would prefer manual counting over the current OGS system.
Score cheating would not be possible. And admittedly there would probably be some extremely rare “wrong” outcomes that could not be fixed.
True, but how is our track record at providing a reasonable UI for manual scoring? Like one where we can know what the markings on the board mean, and what the other player has changed?
Well, unlike the simpletons who think it’s broken because it’s missing wheels with silver hubcaps, the smart people like me know that it’s broken because it’s missing wheels with black hubcaps
I must be missing something (since I’m a mediocre 6k player and might not fully understand the scoring rules for games played against the computer). I generally play against the computer, most recently against Fuego. While trying to figure out the discrepancy in the counting between what I counted and the server counted, I came across a comment about a rule I had never seen before and that doesn’t make sense to me. The rule has to do with the consequence of passing. Do I lose a point every time I pass? Even after counting in the 7.5 points Fuego gets for playing white, I count myself considerably ahead (15 points or so). I started passing when I no longer had any move to strengthen my position (which is obvious, I thought), while Fuego continued to play, generally in territory already fully decided, whether its or mine. If anyone could tell me how this works, I’d be deeply grateful.
Is this the game? geller vs. Fuego
Where did you see this comment you are referring to?
Edit: the predicted score doesn’t really change when you pass. Are you saying you manually counted, and that disagrees with the autoscore?
Bots are usually using Chinese rules. You don’t lose a point when passing under these rules. You don’t lose points for playing in your territory or in the territory of your opponent either.
You lose a point when passing under AGA rules for getting consistency (equivalence) in the scoring (sometimes about 1 point difference) between a territory scoring and an area scoring
In Chinese (and other area scoring) rules, your points are your alive stones plus your territory. So if you pass, this number doesn’t change. So your score as in “Black (you) had 180, white had 181” doesn’t change. The game score aka difference between you and opponent’s version of these numbers may change, depending what they do. If there are still dame (neutral points) available and they play one of these, their alive stones goes up by 1 (and territory remains the same), so you “lost a point” in game score even though your personal score remained the same. There was an opportunity cost, you took 0 points when you could have taken 1 point. If there are no dame and the opponent plays inside their (or your) settled territory, then they gain 1 alive stone and lose 1 territory so their score and the game score remains the same.
Thanks so much for this explanation. I had no idea that the final score is partly a function of territory and partly a function of the number of live stones. So grabbing the dame (neutral spaces) is important. I still don’t understand some of the scoring weirdness of a particular server (Fuego). I would send you a screenshot of the board’s final state, but it doesn’t show the number of previously captured stones. In any event, I’m deeply appreciative of your help.
Thanks very much for your reply to my question. I have to confess that I don’t know the difference between territory scoring and area scoring, but that would only account for a one point disparity (if I’ve understood your explanation), so wouldn’t help me understand how a particular server (Fuego) frequently miscounts by a lot more than one point. I took a screenshot of the final board, but since it doesn’t report the number of previously captured stones, it wouldn’t be very helpful.
You may link the game here so we could have a look and help maybe?
Following the link of a game, you can find the captured stones by hovering over the score and click the ?-mark. There you will see how the score is build up (territory, prisoners, komi).
The poll has ended.
“Manual scoring” is the preferred scoring method for the OGS community.
We should talk feature and feasibility. The issue I see with our winner “manual scoring” is the high possibility of disagreements. We need to make sure that this would not become a GM nightmare.
Ideas voiced in this thread were:
I personally think the best option would be to do full automatic after the game was restarted once. Or alternatively what dexonsmith said.
Additional UI change
This could be seen as a seperate endeavor but is closely related. We should also change the UI so that there is a “proposal” and an “accept” button during the scoring phase. The current UI allows for accidental agreements.
If you implement full auto after game was restarted once, then the issue in this thread needs addressing: Why does 1 pass end game after resuming?
If one player wishes to resume the game, sometimes the game needs to be resumed twice before they can play.
