Re: [gnugo-devel] Owl mistake and thrashing dragons.

[Gunnar Farneback]

Dan wrote:
> Maybe some scheme like this would work. Some patterns
> would be added with an additional classification
> which would mean that under certain circumstances
> they would be treated as 'a' patterns. If there is a
> thrashing dragon, and other mechanisms fail to find a
> suitable attack then shapes would be rerun and these
> patterns would be matched.

A simpler solution could be to have an ordinary 'a' pattern with a
constraint checking that the attacked stone is thrashing. Then it
would suffice to add code for appropriate valuation of a strategic
attack on a thrashing dragon.

> For example at move 222 or 228, some pattern such as:
> 
> Oo?
> ..X
> .*.
> 
> would produce the move at N5.

It's tricky to get the right generality for such a pattern. With lots
of O stones outside the pattern, like below, it's clearly
inappropriate.

.OOX
O..X
O.*.
OOO.

On the other hand a pattern like 

oOo?
...X
..*.
...?

would clearly be too specific. A potential solution would be something
like

Oo?
..X
.*.

:8,Oa

Oo?
..X
.*.

;open_area_size(*)>6

where open_area_size(*) is a not yet invented autohelper function
giving a useful measure of the size of the open area surrounding *.

/Gunnar