Auto scoring

[violaine]

Is the score calculated by a specific program or is it calculated by the “estimate the score” program"?

I guess scoring a finished game is much easier and less prone to failure

[yebellz]

I think I raised a similar question here:

What does the Auto-Score button exactly do?

In the scoring phase of the game, there an “Auto-Score” button. I have been hesitant to try it, since manually marking the stones is easy enough, and I wasn’t sure if that button would automatically commit the user to using the auto-scoring. What exactly this button do? I guess it would attempt to automatically make life/death determinations, but does it also automatically accept/commit the user to that determination? Or are you still able to modify it if something was marked incorrectly?

The auto score button just resets the stone status to the initial position. The players can still edit it and need to confirm it. It’s useful for quickly fixing things if a lot of things get messed up.

Scoring for some bots works differently, where the stone status is automatically decided and cannot be edited by the human player. I suppose this is to prevent abuse. To avoid issues, it’s best to remove all doubt about life and death by using Chinese rules (which most bots already require) and playing on until you’ve captured any questionable enemy stones, and clearly established life for all of your stones.

[AdamR]

I am unsure I understand correctly, but if I do:

AFAIK the estimation during a game is done by the same algorithm as the final count (only maybe with less playouts?).

[violaine]

Hum sorry but It’s not very clear as I don’t speak about autoscore button but how the score is calculated after both agreement.

[AdamR]

online-go/score-estimator

[violaine]

Thanks for the reply. There is no mention of score calculator although in the readme.
So I stay curious, should I conclude the calculator is the estimator?
I will then feel more concerned about all the weird guess from the estimator.

[AdamR]

score estimator AND dead stone removal suggester (remove dead stones being the phase after two passes)

Again, if you are asking if the algorithm is the same for score estimating as well as the final count, as far as I know it is. (there may be a different number of playouts maybe for the final count?) but other than that it should be the same code.

If you are asking about something else please try to rephrase the question I am getting confused

[violaine]

Ok I got it
Thank you.

[Conrad_Melville]

I’m not sure that is right. The score estimator always gives scores using Chinese rules, while the preliminary autoscore during the stone removal phase uses Japanese, Chinese, etc. depending on the game specifications. So at a minimum, there is an addition of some kind to the algorithm used by the estimator.

[violaine]

I was in doubt because other servers with not that better estimator btw (kgs for one but not only) use a different procedure to calculate the game at the end. Which relieve me when I see how deeply wrong the estimator often can be during the game.

But I dunno I really just asked and accept any answer from who knows.

[AdamR]

Okay, seems the question is more complicated that I initially thought

I still maintain that probably basically the same algorithm is used. Going by the github description and this line:

@anoek

If the consensus is that things look better, especially towards the end game, I’ll begin using it for the automatic scoring as well (Score estimator improvements)

But indeed that does not mean the whole code is the same. I am assuming the algorithm is just used for determining alive/dead groups and the scoring itself is run off it. But more I rather dare not discuss, we can involve anoek if it is somehow important

[flovo]

@violaine

Is the score calculated by a specific program or is it calculated by the “estimate the score” program"?

I guess scoring a finished game is much easier and less prone to failure

@violaine

Hum sorry but It’s not very clear as I don’t speak about autoscore button but how the score is calculated after both agreement.

After both player agreed which stones are dead, nothing (fancy) happens.

In the stone removal phase, you can see which empty point is scored for which player (the white/black squares). The scoring algorithm just counts this points as territory, so this is almost fail-save.
The dead (transparent) stones are added to the prisoners and counted as such.

In the stone removal phase you can see how many points each player have. This score will not change after both player accepted it.

I don’t know what algorithm is used for the dead-stone-detection after both players passed. But after this initial process the server only counts what both players decided is dead/alive and assigned as territory of black/white.

---

The problem of the score estimator is the dead or alive detection and the assignment of empty points as territory. It shows its assignments by using black/white squares. Counting of this points is something even the SE gets right

[violaine]

Maybe I’m unlucky but the few times I was curious of the estimation, I had to rescore by myself. Not only because of uncertainty still not clearly fixed (I do understand) but much worse because he marks dead as alive chains of stones (or reverse ofc) Seems to have difficulty with bamboo joint for example.
Gladly the final scoring of the game looks working much better. So somehow I was wondering.

[flovo]

Just to be sure, you know you can change the live/dead status of stones in the stone removal phase by clicking on them. The same for territory points in case Seki is detected wrong.

[violaine]

I have no difficulty on this side, thanks for your concern. Tricking the abusing players who even don’t (or can’t) check the estimation may be the point to keep it running like this?