From the comments in that thread, it seems there was no dispute about the the left side. Both players seemed to agree that white’s 2 stones on the left were simply dead and this agreement is sufficient to settle the left side as black’s territory.
1 Like
As gennan said, if both players agree the 2 white stones are dead, then they are removed from the board as dead stones during the counting phase which means the black group no longer has an dame so is alive with territory rather than alive in seki with no territory.
The problem here is that the players ended the game when there is still a valuable move left on the board which they overlooked due to lack of skill, and the AI powered score tool/estimator informs them of it. Japanese rules can deal with it, from http://www.cs.cmu.edu/~wjh/go/rules/Japanese.html:
9.3. If a player requests resumption of a stopped game, his opponent must oblige and has the right to play first.
Trouble is the opponent going first means you want them to be the one to request a resumption so you can play the good move. This is dealt with by saying if they can’t agree to end the game, they both lose:
13.1. After the game stops according to Article 9, if the players find an effective move, which would affect the result of the game, and therefore cannot agree to end the game, both players lose.
No Go servers I know of implement this. A decent referee would.
4 Likes
Wow, this is interesting. Thank YOU.
It seems obvious that in this instance both players determined the left side of this board for white. Then the scoring tool reversed the TACIT agreement between the players.
It’s mechanized kibitz and it comes in at the exact worst time.
Finegold? Think I’ve seen it.
1 Like
Indeed Finegold.
But in this case it’s just the trivial counting part that reversed it, not the more controversial AI auto-scoring algorithm that automatically marks dead groups. Do we need to let players miscount seki as territory?
1 Like
I don’t understand. The tool wouldn’t allow them to count that territory as white, because it was in settled to 9d scoring AI. (And certainly lower rank observers). To those players that left side was white.
In this case the problem is the players not marking the territory manually as white and then both agreeing to that.
I guess there is the issue that the automatic scoring is seen as since kind of authority that can’t be questioned.
The pink region in your screenshot can’t count as white, since the black group on top is alive:
No 9d AI required; territory just can’t be touching live stones of the other color.
3 Likes
But you could mark everything but e9 and d9 and white territory and not be far wrong…
1 Like
This is just a player mistake that makes it impossible to count that territory as white’s.
How would you score a 9x9 game where both novices pass while both have 40 stones on the board and 1 shared liberty?
Or IRL games where both novices pass while there are multiple strings of stones on the board with 0 liberties?
At some point you just have no other choice than give up and make no further attempts to come up with a result that seems reasonable to everyone.
1 Like
I agree with those people who’ve said that scoring is part of the game.
If Black thinks his group is dead and marks it dead, and White believes the same, I don’t think the actual reality of whether or not it can live is relevant.
4 Likes
From what I’ve been linked to in another thread, something like Tromp-Taylor Rules at Sensei's Library seems to score any final board configuration 
But you would be very far wrong, for scoring this position. You cannot mark individual intersections as territory, only contiguous regions. And territory is a continuous region surrounded by stones of only one colour (after removing dead stones). The left side region is surrounded by mostly white but also black. What your are describing is the answer to the question “given competent players continue this game to a sensible conclusion, how much of the left side will be white territory?”. This is also the question score estimators attempt to answer, with the new KataGo powered one assuming extremely high skill rather than just competency. But this is a different question to scoring this position when the players decide not to continue. This is one reason I think score estimators are a bad idea (at least for beginners without warnings and guidance from a teacher), because they mislead beginners info confusing these 2 questions.
When I teach intelligent university students the rules of Go in person, they will generally understand they need to close the boundaries of territory after the first session. 10 year old children can take longer. That someone can be 13k on OGS and not know this surprises me. It indicates deficiencies with how people learn the basic rules of go online.
1 Like
No, the counting tool was simply enforcing the rules of counting. A region can’t be territory if it is adjacent to both black and white stones. Counting tools have done this for decades because it’s a simple graph connectivity problem that would be a reasonable computer science homework problem 50 years ago. Automatically detecting dead stones is a significantly harder problem, the old OGS score tool, or the KGS or other server ones do an okay job in simple cases but can get it wrong, whereas the new OGS KataGo one is much “better”.
The players can’t agree to break the rules of the counting phase of the game, just like they can’t agree to leave stones with no liberties on the board (in person you can, but it’s the job of server software to enforce the rules).
I didn’t check how many games were played by this player but isn’t part of the problem here that you are effectively 6k or something a long way from 25k when you first join OGS for algorithmic purposes (modified by humble rank blah blah) and then it takes a while to settle now. Especially if you are catapulted up the ranks by early resignations by strong players or matchups like this one where two over ranked players meet.
I think the OP has played enough games to know that borders need to be closed, so I assume this case was simply an oversight by them.
I have made this oversight once or twice myself on servers that give no feedback about territory status while marking dead stones (areas being fully enclosed + only seeing opponent stones that are marked dead).
1 Like
I did check, and they have about 100 games. I quickly looked at a couple of games and whilst the play is weaker than I’d expect from an EGF 13k, it seems plausible as a 13k in new OGS rankings, and it was a lot stronger than a 28k beginner I have just taught and has understood about territory borders needing to be finished.
1 Like
The border at E9 is not closed so the scoring ai does not recognize / admit that the left side is white’s territory. Actually, white is ahead. Anyway, it seems the game was annulled so no win / loss.