I understand what you are saying but I seem to be doing a bad job of explaining my point.
Being more precise assumes that there is something worthwhile to measure with that extra precision. Otherwise, it is just injecting noise into the score.
I don’t see why we should assume territory is the primary unit and area isn’t. I’m asking why that assumption is justified.
Taking the highlighted row of that table, I don’t understand why, in abstract, we should distinguish between black surrounding 6 points of territory and getting one extra space to play a stone as opposed to black surrounding 7 points of territory and thus not having room for the extra stone.
He got 7 area in both cases. We can split that area up in a other ways that also create parity differences. For example we could have area on the 4th line and below vs area in the center, there will again be two different parity cases for that split that lead to the same total area.
By saying that territory is more precise because it distinguishes these cases, you are assuming that there is something important that should be distinguished.
I want a non-circular explanation of what that something is.
If we think of goal as creating space for your own stones by establishing area where the opponent cannot play, then a “complete” playout would fill the board with stones and single eye spaces. That gives an area score.
So, in some sense, territory is “unnatural”. That doesn’t mean it is wrong, just that it requires some justification to explain why we think this distinction is an important matter of skill.
Yes, there are cases where things we think of as “mistakes” are penalized by Japanese rules and not by Chinese ones. But there are also cases where Chinese rules punish a mistake that Japanese rules do not.
Is there a reason to give priority to one set of situations over the other? (Though, perhaps area scoring with a parity correction captures a good mix of both. I haven’t thought carefully about it.)
The most intuitive argument seems to be that we want “equal resources” and so we compensate the player who used fewer stones by the difference in the number of stones each player used.
But it assumes that there is no real skill involved in a player winning that extra stone, but there is in winning the extra territory. It also seems to assume that the sequences that lead to these outcomes are somehow meaningfully different and shouldn’t be miai.
You could argue that, since it’s a perfect information game, even though we don’t know which way is which, we do want to be able to tell the difference. But then you have to give up the ability to tell the difference in other cases. And it goes back to “why this difference and not those?”
Perhaps, even if the two situations should be miai in theory, arbitrarily perfering one over the other makes it easier to evaluate a position and find good plays. So in practice, with human players you get a more entertaining game.
If that’s the case, I would think that would translate into an AI using territory having less uncertainty (more precision) with positional evaluation. And a steeper gradient for the losses in value caused by suboptimal moves.
That said, in abstract, you can also say that this is undesirable because reducing the number of viable plays limits creative strategic opportunities.
So, it’s not a matter of just asserting that territory is more precise. It’s asking why we care about that precision and what it is we are actually measuring to justify the extra complexity.
The best argument I’ve seen in this thread is the pragmatic argument that at a high level, under territory rules, a komi set to give each player equal winning chances will also have fewer draws. That’s pragmatic and a matter of taste, but I think it fits with the general ethos and history of the game to want to avoid ties.
