NAME
Game::Go::Scoring - Japanese territory, and the area score it is checked against
VERSION
Version 0.01
SYNOPSIS
my $result = Game::Go::Scoring::score($game);
say $result->stringify; # "White wins by 6.5"
Game::Go::Scoring::agree_under_equivalence($game); # undef, 1 or 0DESCRIPTION
Japanese territory scoring, Articles 8 and 10, in five steps:
1. remove every chain agreed dead, adding each stone to the prisoners of
the player who did NOT own it (10.1)
2. flood-fill the empty regions of what is left
3. a region reaching exactly one colour is that colour's eye points;
a region reaching both is dame (8)
4. territory is the eye points, less any region agreed seki (8)
5. black = territory(black) - prisoners held by WHITE
white = territory(white) - prisoners held by BLACK + komi (10.2)Step 5 is subtraction from the opponent, not addition to yourself. Article 10.2 fills each player's prisoners "into the opponent's territory". Adding your own prisoners to your own score gives the same difference and a different pair of numbers, and the pair is what a scoreboard prints and what an SGF record's RE[] encodes.
Where this scorer is knowingly incomplete
Article 8:
Empty points surrounded by the live stones of just one player are called
"eye points." Other empty points are called "dame." Stones which are alive
but possess dame are said to be in "seki." Eye points surrounded by stones
that are alive but NOT IN SEKI are called "territory," each eye point
counting as one point of territory.The British Go Association's comparison table makes it the row that separates Japanese from every other ruleset on the page: "Can you count points in a seki? Japanese, Korean: No. AGA, Chinese, SST (Ing), New Zealand: Yes."
A flood fill identifies dame correctly, because a region reaching both colours is dame by definition. What it cannot identify on its own is a one-point eye inside a seki: it reaches one colour, it is not dame, and it is still not territory, because the stones around it are in seki. Seeing that requires knowing which groups are alive-with-dame, which is life and death again, and this distribution has no solver.
So seki is agreed, not detected. The confirmation phase lets the players mark a region as seki, and a marked region scores nothing. A position where the players do not mark it will be scored one or two points wrong in exactly that shape, t/14-seki.t asserts that wrong answer as documented behaviour, and the escape hatch is Article 9.3: a player losing a point to it can send the game back to the board and play the seki out, which is what a human referee would say.
FUNCTIONS
score
Game::Go::Scoring::score($game, scored_by => 'territory')score_by_area
Game::Go::Scoring::score_by_area($game)A Game::Go::Result scored by Tromp's rule 9: "A player's score is the number of points of her color, plus the number of empty points that reach only her color."
It reads no marks, and that is the whole reason it is what a disputed game falls through to: it is a function of the position and of nothing else, so nobody has to agree to anything for it to produce an answer. Komi still applies, because it compensates for moving first and that is true under either ruleset.
raw
The scorer's own numbers, as a hashref, before they are dressed as a result. Scores in it are in tenths, because komi is fractional and the C engine has no floats.
stones_played
How many stones each colour played over the whole game, from the log. Not the same as the stones each has on the board; the difference is exactly the prisoners, and that identity is what the next function rests on.
agree_under_equivalence
undef the condition does not hold
1 it holds and the two scorers agree
0 it holds and they DISAGREEThis is the differential test that a single implementation of everything else would otherwise leave this distribution without. There is no pure-Perl board to compare the C against and no published ladder to check it with, but there are two scorers over one position, and they must agree under a stated condition.
Wikipedia's "Rules of Go":
If the game ends with both players having played the same number of stones,
then the result will be identical in territory and area scoring.The condition implemented is equal stones played, and no region agreed seki, which is derived in a comment in the source rather than taken on trust. Two things that look as though they should matter do not: removing dead stones preserves the identity, because a lifted stone moves from the board count into the prisoner count on both sides of it; and unfilled dame preserves it, because dame counts zero to both scorers. An agreed seki does break it, because territory excludes a seki and area does not.
A 0 from this is a bug in one of the two scorers.
SEE ALSO
AUTHOR
LNATION <email@lnation.org>
LICENSE AND COPYRIGHT
This software is Copyright (c) 2026 by LNATION <email@lnation.org>.
This is free software, licensed under:
The Artistic License 2.0 (GPL Compatible)