AreaScoring/ Discussion

Area scoring is nice conceptually because it is simply about dividing the points on the board between the two players, Black and White. If every point on the board was always given to either Black or White, the score would always be an odd number, i.e. White wins by 1, or Black wins by 5. Because of seki, however, not every point on the board is always awarded to either player. This can lead to situations where the score is not an odd number like usual, but an even number instead. Perhaps it is worthwhile to consider other definitions of the score which always assign each point on the board to either Black or White. This might be thought of as conceptually nicer Also it makes the scoring a bit more regular, and maybe this is a good thing.

-- The Count

Possible Score Definitions

1.

A point counts for Black if Black's closest stone is closer than White's closest stone, otherwise it counts for White.

This is mostly the same as normal area scoring except White gets both neutral points in a normal seki.

How "close" a stone is could be the manhattan distance, or some other measure.

2.

A point counts for whoever has a stone closest to it (see proximity scoring). EXTENSION to proximity scoring (or EQUIVALENT?): if both players have a stone equally close, remove those two stones (imaginarily) and start again.

This is fairer, but on a completely symmetrical board, a point may still be undecided. A number of variations on the rule could decide this point.

This still does not agree with normal area scoring in many seki positions. I've since found out that apparently Ing rules used to award points in seki in a similar way.

3.

After removing all dead stones, the players start playing again, filling empty points that are reached by both colors (shared liberties in seki). Stones are no longer captured when surrounded, but remain on the board. The first player to pass moves first. This way, shared liberties in seki end up acting like extra dame points --Herman Hiddema

Are there better potential definitions out there? Please comment.

Skylancer: I prefer the standard definition. A point is awarded to a player if that point is either occupied by the player's stone, or part of a region completely surrounded by that player's stones. While it may leave neutral points on the board in sekis unawarded, it is simple to implement and use. I don't see a need for assigning every point to a player.

The Count: I don't think I'm biased in saying that I think they're both equally simple.

Besides, in completely symmetrical positions with leftover neutral points in seki, it seems arbitrary to declare that a such a point be awarded to one player or the other. After all, the position is symmetrical! If neither player can safely fill/surround and claim such a point, then rightfully, neither should be able to count it as part of his/her score.

The Count: Yes, I think the symmetry thing's the main problem. One solution is to not award anyone the point if it's really equal. Maybe this goes against the original idea. Note though, that in usual area scoring, a komi of 7.5 can be regarded as a komi of 8 where Black is given the odd neutral point if there is one. Ing rules essentially do this by saying komi is 8 and Black wins ties.

Now, there would be a use in finding an alternative score definition for computer-Go mid-game score estimating, since rarely are any regions completely surrounded mid-game. But if this we're shooting for, we might as well go all the way and add in a life-death-evaluator and an influence heuristic and everything else.

axd: these are all needless complications due to lazyness, leading to problematic positions. just fill in the areas as the ruleset proposes (ie agree, or disagree and follow the procedure), and there should be no (or hopefully far less) problems.