[gnugo-devel] bogus vital attack pattern VA18

[Martin Holters]

Hi all,

the following pattern looks a little fishy to me:

Pattern VA18

XX!O       throw in to kill an eye
.*X?
XXO?

:8,-,value(35)

XXaO
.*X?
XXO?

; oplay_attack(*,a,a)

Shouldn't this be classified as 's' to make sense?

I've run the regression suite with this pattern removed and with it
classified as 's'. In both cases, no delta except for the node counts,
where removing the pattern slightly reduces the reading nodes, and
adding 's' slightly increases reading and owl nodes and reduces
connection nodes. Increase of the owl nodes is most noteworthy with
0.012%.

// Martin