Okay, I know I made some mistakes and got back from them in the middle of the game, but why does White ultimately resign? The estimated score for the board in the state it ends is White by 0.5, and I’m not sure where I could swing even one more point in my favor. Please tell me what I’m missing!
Well, it looks like you were ahead.
While the score estimator does say that white is ahead by .5 points, this assumes you’re playing under Chinese rules which is 7.5 points Komi. However, you guys were only playing with .5 points komi, so there was no way for your opponent to win.
At least… that’s what I think.
The situation is more confusing than it first appears (which is why I deleted my initial post).
I experimented by forking and challenging a bot (Master Mantis), with me as white.
Then I just passed. Master passed also and the result was White by 1.5.
But wait, there’s more.
Then I did it again, but I filled the one dame I could find before passing. Master responded by also filling, then we both passed.
This time the result was White by 0.5.
How can that be, under Japanese rules?
I was so suprised I did it a couple of times just to make sure, before I thought of checking my game history. The tests are in my game history.
Japanese sure doesn’t work the way I thought it did!
In fact, I can see what appears to me to be a problem: at the end of this game, A8 is being counted for white.
EDIT: Sorry @GreenAsJade , I just saw your second post. The score estimator does its best to decide the correct points. It’s up to the human to make sure everything is correct. This is a known problem, not a bug, but more of an inconvenience due to the fact that it’s hard for a computer to count… I think.
I took a look at your games. The reason for the discrepancy appears to be because when you didn’t fill in the Dame, you allow A8 to be counted as a point for white, which it isn’t.
I hope that helps.
I’d say more just an inaccuracy within the score estimator… none of them are perfect.
And you beat me to it.
However, regardless of whether white wins by 0.5 or 1.5… OP’s question of why white resigned is still a confusing one… simple error of judgement perhaps? misread?
Yeah… I’ve got nothing.
I think you guys are missing my point.
The final result is not “the score estimator” it is “the final result”.
It appears to me that in the case where I didn’t fill, the final result is incorrectly calculated. If black had one more point elsewhere, this would make the difference between winning and losing.
Isn’t that a bug?
A8 simply should not be counted in the final result.
Wait, I just reread this:
Are you saying that after a game finishes, I have to count manually because it might be hard for the computer to count, and that the computer’s calculation of the final score is not necessarily correct?
I think I understand.
Both the score estimator and the final score count A8 as a point for white. Which should make sense because both the point estimator and the final count should be using the same algorithm.
It’s something very common and more of an imperfection than a bug. Due to the limits of trying to program Go computers.
That of course, is why the final decision for the score is up to the human players and not the computer.
Not the entire score… but this is what the score review stage is for. You approve what stones are dead and alive and if you encounter any errors such as these you go back and play them out to correct (if it’s important)
Oh I could have won? I was worried about the top-right group, but I guess even with it dead I was barely ahead. Maybe I should start using the score estimator.
Top right group is dead but even so I think passing and going to score review is better than resigning. Black has no good invades so just pass instead of resigning <3
Ah - I see, my mistake was not to correct A8 in the removal phase: I get it now
Which emphasises the point I made in the deleted post That is that people way over-estimate the accuracy of the score estimator! In this forum, we’ve seen people feeling sad about losing when they are resigning early in the mid game based on a losing score estimate! Likewise, this thread shows us that if the game is close in the end game, the estimator is similarly suspect.
I find the latter more challenging, myself. I can understand that the estimator is way off in the mid game, but I would hate to have to add up to the exact number myself to know if I’m behind or head in the end game
Luckily at my level games hardly ever seem to go down to the wire in that way
I appreciate all your input! I mostly use the score estimator to see if it’s as close as I think at the end, or to double check areas that I think I control during the middle of the game. I wasn’t 100% sure of its accuracy, but it seemed to be an awfully close position, even with the top-right group. Thanks for the analysis of how accurate it may or may not be – that helps a lot.