Re: [gnugo-devel] score_3_16.1

[bump]

I wrote:

> But what is the exact connection between this problem in score.c
> and the unbalanced results you are noticing? Do you have an
> example of a game that is misscored?

I too have noticed that B tends to lose more games in
twogtp series, and that the right komi for GNU Go is
closer to 8 points than to 5.5.

Notice that the suspicious block of code is inside a loop
conditioned on critical dragons being found.

Presumably this error has no effect on the scoring of games at
the end of the game because there will be no critical dragons at
the end of the game. So misscored games cannot account for this
problem.

It is possible that it causes black to lose more games because B
thinks he's behind when he's not when there is a critical dragon
on the board, hence does not play cautiously.

Dan