[gnugo-devel] An owl problem

[Daniel Bump]

Consider the situation at move 70 of the game:

regression/games/nngs/looknglass-gnugo-3.1.26-200203042158.sgf

GNU thinks the W dragon around N19 is dead. It looks like
this:

   A B C D E F G H J K L M N O P Q R S T
19 . . . . . . . . . . O . O O X X . . . 19
18 . . O . . . O . . . X O . O O X . . . 18
17 . X X O O . O . X . X X O O X . . . . 17
16 X . . X X X . O . + . O X X . X . . . 16
15 . X . . . . . . . . . . O . . . . . . 15
14 . O O . . . . . . . . . . . . X X . . 14

Clearly a white move at K19 or K18 (or for that matter J19)
would save it, and in fact the pattern D1374 is matched.

Pattern D1374
# gf New pattern. (3.1.14)

oo..Xo        connect underneath
...*O.
------

:8,-,value(55)

ab..Xo
...*O.
------

;o_somewhere(a,b) && owl_escape_value(a) + owl_escape_value(b) > 0

Unfortunately, it is not perceived that this move connects out.

One idea is that the goal dragon should be expanded to
include the stones around G18. Another idea is that
since the stone at G18 has owl escape value 6, connecting
to it should be considered a win.

Actually these ideas could be unified if we introduce
a new class of patterns in owl_defendpats.db. This
classification (say class_C which is unused in that file)
would be given to pattern D1374.

Any O stones outside the goal matched in a C class pattern
would be added to the goal dragon. And if such a stone had
owl escape value 6, the pattern would be treated as a win.

Dan