Scoring question

I just a game where my opponent scored a lot of points for blank points which were not surround, and am confused as to why. Thanks!

Could you provide a link to your game? The last games that were played with the account you are currently commenting with, are from 2021, or did you not play on OGS?

Thanks for your reply!

Here is a link:

I was white, and the black player scored a lot of points (147 I think) for the intersections in the entire corner section of the board.

Ah sorry, I was only looking at 9x9 games for some reason.

This game is technically speaking impossible to score, since there is a lot of intersections that are not “territory” for either player.

If we look at the empty space on the outside, it is connected to both the white stones and to the black stones, and both the white group of stones and the black group of stones are alive. In order to be scored correctly, empty space needs to only connect to either

  • (living) Black stones and White stones that are dead (i.e. White cannot prevent capture)
  • (living) White stones and Black stones that are dead (i.e. Black cannot prevent capture)

In this case, the outside empty intersections do not fit either description. Therefore it can technically not be scored. The automatic scoring algorithm of OGS has therefore made a guess and assigned some (more or less) arbitrary points to Black.


To properly score the game, both players need to close their territories by playing stones across the entire border. For example, in the following position, where I continued the game to close the territories off, the territories are defined properly:

All the Black territory is marked with a square, and all the White territory with a triangle. Note that E13 is marked as dead, and so it belongs to the White territory.


See also this page for more information on counting.

4 Likes

Why is the autoscore scoring a huge number of unsurrounded points (i.e., with open boundaries) on the right, but none on the left? This is reminiscent of what the autoscore used to do in 2021, before the revision, but even then, the system would have interpolated points on both sides. This is very bizarre and needs an explanation.

1 Like

I guess the explanation is that the autoscore works in mysterious ways.

It has certainly been improved since 2021, we get much fewer reports of wrongly scored games, and almost all of those reported for wrong score are due to a player marking stones wrongly, and not due to autoscore.

I can imagine that the extreme openness of the mentioned board was what caused the autoscore to mess up.


That said, no scoring algorithm could’ve scored this game meaningfully: it is not ready for scoring.

1 Like

Using my skills as a KataGo whisperer (and without actually doing any experiments to ask it), I suspect it’s related to the single black stone at e13 (or the black group further down on the left), versus the lack of any white stones on the right.

1 Like

Thank you for the explanation!

1 Like

In case anyone wants to try to make sense of the weird display, here is a predicted “ownership” from KataGo (copied from a request to https://beta-ai.online-go.com/api/score) that I think is the kind of thing autoscore uses:

[
[-0.263,-0.252,-0.28,-0.344,-0.451,-0.557,-0.646,-0.701,-0.714,-0.688,-0.556,-0.437,-0.25,-0.104,-0.018,0.042,0.078,0.099,0.127],
[-0.254,-0.237,-0.258,-0.344,-0.484,-0.607,-0.694,-0.755,-0.809,-0.807,-0.804,-0.477,-0.276,-0.111,0,0.058,0.105,0.117,0.135],
[-0.268,-0.201,-0.114,-0.312,-0.418,-0.557,-0.643,-0.754,-0.833,-0.888,-0.993,-0.594,-0.207,0.085,0.034,0.108,0.154,0.134,0.151],
[-0.307,-0.361,-0.256,-0.321,-0.453,-0.604,-0.746,-0.998,-0.999,-0.999,-0.98,-0.991,0.036,0.091,0.14,0.122,0.05,0.167,0.184],
[-0.352,-0.407,-0.337,-0.387,-0.539,-0.999,-0.999,-0.993,-0.994,-1,-0.985,-0.993,0.902,0.43,0.25,0.231,0.229,0.288,0.247],
[-0.356,-0.393,-0.368,-0.412,-0.607,-0.999,-1,-0.995,-1,-1,-0.999,0.971,0.861,0.996,0.465,0.346,0.332,0.351,0.327],
[-0.325,-0.354,-0.326,-0.411,-0.279,-0.999,-1,-0.997,-1,-0.997,-0.999,0.976,0.955,0.998,0.64,0.448,0.41,0.43,0.394],
[-0.275,-0.309,-0.332,-0.458,-0.977,-0.963,-0.992,-0.997,-0.997,-1,-0.999,-0.995,0.997,0.981,0.993,0.561,0.475,0.463,0.443],
[-0.183,-0.218,-0.253,-0.39,-0.609,-0.999,-0.999,-0.996,-1,-1,-0.998,0.99,0.982,0.998,0.75,0.584,0.518,0.507,0.481],
[-0.048,-0.089,-0.144,-0.305,-0.514,-0.999,-0.993,-0.999,-0.999,-0.999,0.999,0.985,0.999,0.804,0.667,0.613,0.56,0.539,0.51],
[0.128,0.031,-0.145,-0.231,-0.534,-0.999,-0.999,0.989,-0.996,0.999,0.999,0.99,0.999,0.747,0.678,0.631,0.592,0.57,0.538],
[0.341,0.289,0.083,-0.017,-0.947,-0.932,-0.999,0.996,0.998,0.993,0.994,0.999,0.812,0.706,0.676,0.642,0.612,0.599,0.569],
[0.629,0.547,0.443,0.964,0.962,-0.998,-0.999,-0.996,0.998,0.999,0.992,0.999,0.746,0.684,0.669,0.661,0.635,0.632,0.603],
[0.821,0.976,0.809,0.961,-0.992,-0.998,-0.998,-0.321,-0.374,0.998,0.985,0.999,0.699,0.647,0.648,0.658,0.674,0.667,0.645],
[0.884,0.912,0.981,0.644,0.667,-0.995,-0.911,-0.943,0.726,0.712,0.993,0.718,0.599,0.595,0.594,0.641,0.813,0.767,0.639],
[0.832,0.837,0.84,0.959,0.214,-0.441,-0.969,-0.406,0.208,0.967,0.655,0.525,0.527,0.521,0.537,0.406,0.856,0.561,0.592],
[0.749,0.739,0.681,0.612,-0.027,-0.081,-0.428,-0.124,0.058,0.47,0.416,0.446,0.443,0.436,0.443,0.489,0.14,0.573,0.478],
[0.684,0.623,0.533,0.315,0.062,-0.184,-0.204,-0.171,0.112,0.191,0.347,0.383,0.398,0.407,0.359,0.308,0.272,0.396,0.444],
[0.597,0.512,0.398,0.204,-0.012,-0.148,-0.2,-0.129,0.006,0.172,0.27,0.326,0.342,0.323,0.306,0.279,0.328,0.363,0.409]
]

Looks like the blue squares, whatever they are supposed to mean, correspond to ownership of magnitude less than about 0.124.

Let’s see about that :wink:: https://pdg137.github.io/autoscore/v3.html?game=54084826

Since the players are done playing, all stones should all be marked alive, and the entire outside region is dame. White and black have 16 and 11 points of territory, respectively.

It is mutual resignation.

1 Like

I really disagree with this statement.

Any game in which both players passed is finished, and any finished game can be scored. The correct result would be to consider it all as dame.

That being said, @dorsata why did you pass?

2 Likes

I think this is only the case if both players agree on which stones are dead.

1 Like

Right - I should have said “both players passed and agreed on the status of their group”.