I can’t speak for my opponent exactly but I will assume. I thinks it’s fair that we both thought the territory on the left belonged to black. In an IRL game we would have both counted it as black and that would be that.
It’s my opinion, that the scoring tool in certain cases is like a pro player telling you or your opponent, “you can kill that group and win”. A type of kibitz that is unacceptable in all competitive games.
I’m appreciative of the developers and community. It’s not meant as criticism, just my perception.
I know that the time I restarted and beat my opponent using the info from kata go, I personally felt like a cheat afterwards. Had my opponent in this game done so or had the difference of Kata killing the group made me lose I would have been really bummed.
I don’t really get which solution you are proposing in case players pass too early.
Do you want players to always mark dead stones by themselves?
Or do you expect KataGo (or some other software) to guess correctly which status evaluation of groups the players seem to agree on (without giving them hints about the actual status of groups)?
I’m not going to rewrite everything in the linked posts, especially not Vsotveps steps.
All I said in my case was to mark stones that were pass alive as points for that player with Katago, and similarly with areas that don’t change territory color with a pass.
It’s not a good system in and of itself, only a first step really.
In the OP’s case “too early” was not recognising aji which is quite different to the Reddit example you linked where most of the borders aren’t closed, and yet still the first step would work the same.
(1) I see some people don’t like that automatic Scoring scores game more correctly than players.
(2) But, when opponents do not agree with each other, they often start asking questions in English chat - and then entire chat starts to give them advice. And later moderator may choose winner instead of opponents.
As far as I know, it is the same in real life tournaments; If players can’t get to a game conclusion, then the arbiter is needed to settle the discussion.
Currently there is problem that some players don’t accept score for indefinite time.
So I think by default Scoring should be slightly more stupid - and there should be text that “status of groups may be incorrect, change it if needed.”
And analysis mode should be available - so players can show proofs to each other without returning to real game, its especially needed in Japanese rules.
If players don’t agree with each other for 5 minutes, full power Kata auto-score should be enforced to choose winner.
If players still not ok with it, they may call moderator.
Players should not be able to indefinitely increase time of waiting of score - then that trolling problem will be finally solved and winner will be correct in most cases.
What if AI scoring is just an optional tool, players can request via a separate button on the scoring screen? (“request AI scoring assistance”)
With a fair warning, like “A superhuman AI will now evaluate your game. This could reveal important information to you or your opponent, which could change the outcome. Are you sure you want this?”
Ai scoring only shows up only when one of the players requests it, and the other accepts. If not then they can agree, or play it out. (Or ask help in the chat, call a moderator, etc…)
Normally, if there is no dispute, players just score manually, and only ask for the AI scoring aid in rare problematic cases. This way no game would be spoiled by the AI without the players explicitly requesting it, knowing the risks.
I think experienced players have no problems with the few clicks it takes to mark the dead stones, also they have a pretty good idea of what is dead or not - they have just played through the endgame, certainly if they wanted to live or kill then they would have played some moves to signal that.
Of course there could be some confusing seki situation, that neither of them recognised until the scoring, so there is still the option of calling in the AI.
Beginners who have no clue can just ask for the AI to automatically score the game (much easier than to find someone in the chat). And they also learn that the scoring is actually part of the game, which should be possible without the extra aid, and at some point they will not even need the AI to mark their dead stones for them - but until then, the help is just a few ciicks away.
I am honestly not understanding the issue. The scoring tool works much as it ever has, so far as I can see. It makes guesses about life and death and scores appropriately, as an aid to the players.
Then the players get their chance to click on groups to change their life status, since sometimes AI may fail at this. Only when both players agree that the score is correct should they click the agreement button. If they disagree, they can always click the “call moderator” button and wait for a human decision.
Yes, triple sekis or triple kos can be difficult for a newcomer (sekis are always alive, though). And, depending on the scoring type that was chosen, there can be some obscure cases. But I just don’t understand blaming the scoring tool for giving information.
If that helps one of the players, then that player likely overlooked something, which is human.
If a game truly is won or lost because of, or somehow during scoring, that reveals that the players don’t know how to play go and should read a good go book. I have 25 go books, and almost all of them assume the reader knows how to do scoring.
If there is an issue here, can someone explain it clearly?
Normally players themselves decide what score is. If I have no idea how to kill group of opponent that I surrounded, therefore its not my territory, even if its possible.
Referee tells his opinion only after, if players can’t decide themselves.
Now:
Artificial Referee with above 10 dan level tells his opinion before opponents decide.
Big fun part of game was completely removed.
I agree, this may be the best way. It doesn’t take much longer, and avoids AI interference.
The issue is that beginners may overlook some dead stones, either (1) by being too quick, or not understanding the process, or (2) because they don’t realize it’s dead.
(1) doesn’t worry me too much. You could have an informative message saying that you should now identify dead stones and click on it, and people will get the hang of it quickly (and you can count on the losing player to be thorough when looking at the board, if it’s close).
(2) is working as intended. If the players don’t see a stone as dead, then it’s not dead for the purpose of their game (even if an AI could kill it).
Currently interface of marking groups works weird, even with ctrl button. It should be improved first. Then score estimate tool also will work good without bot opinion.
But, there is no sense in clicking on every single group of stones. All we actually need is strong bot that always make errors on purpose. So we wouldn’t be able to trust it anyway - and wouldn’t need to do many clicks.
no, real solution. Most of the board will be automatically marked correctly, but it will be impossible to know which part. So in fact we are not getting outside information and have to check everything ourselves. But wouldn’t need to do many clicks.
If we’re going to talk about the true score, how should OP’s game actually have been scored? Reading the Japanese rules pedantically, it seems like most of the left side of the board (all regions connected to the dead group) should count as seki, meaning White should have lost most of their territory. Is that really how it works? If so, discovering a dead black group could sometimes result in a gain of points for Black!
If my opponent and I had been playing heads up we would have counted the red area as black. If anyone thought the top part was dead they would have attacked it.
