Hi,
For this question, I am assuming Ray is playing according to Chinese rule (area scoring). Please correct me if I am wrong.
There is not 100% consensus regarding area scoring and handicap games, but it is usual to add the number of handicap stones to the points made by White. So that White score, after removing all dead stones, will be the sum of :
- the number of empty points surrounded by white stones only
- the number of living white stones on the board
- the value of komi (usually 0.5 point for handicap game)
- the number of handicap stones
For instance, Leela Zero accounts for it:
https://github.com/gcp/leela-zero/blob/5958e02cf539d42ee20704b664043036d5c3c18b/src/FastState.cpp#L150
And one can make the assumption that Leela does the same.
GNU Go does the same as well (http://git.savannah.gnu.org/cgit/gnugo.git/tree/engine/aftermath.c#n1185).
From what I can read in the code; Ray does not account for it:
|
score = UctAnalyze(game, S_BLACK) - komi[0]; |
Could you confirm?
Hi,
For this question, I am assuming Ray is playing according to Chinese rule (area scoring). Please correct me if I am wrong.
There is not 100% consensus regarding area scoring and handicap games, but it is usual to add the number of handicap stones to the points made by White. So that White score, after removing all dead stones, will be the sum of :
For instance, Leela Zero accounts for it:
https://github.com/gcp/leela-zero/blob/5958e02cf539d42ee20704b664043036d5c3c18b/src/FastState.cpp#L150
And one can make the assumption that Leela does the same.
GNU Go does the same as well (http://git.savannah.gnu.org/cgit/gnugo.git/tree/engine/aftermath.c#n1185).
From what I can read in the code; Ray does not account for it:
Ray/src/Gtp.cpp
Line 520 in e5b2cad
Could you confirm?