Re: [gnugo-devel] trevor_1_14.5

[Gunnar Farneback]

Dan wrote:
> Here's the pattern:
> 
> Pattern A1
> # sometimes sagari is better
> 
> ?.*Ox        connect
> ..OXX
> .....
> .....
> -----
> 
> :8,n,value(80)
> 
> If O neglects to reinforce X can substantially improve
> his eyespace in sente by hane underneath. Usually he can
> push at least twice in sente.
>
> I'm surprised at this opinion. How would you make it more specific?
 
The comment to the pattern is important. If the purpose is to limit
the X eyespace on the bottom, descending or hane is much more severe,
unless there's some serious problem with the cutting point at *. In
fact we don't want to spend time and nodes on playing * at all if the
cutting point is non-problematic.

I wouldn't be surprised if we today have patterns elsewhere to detect
problematic cutting points, which would find the move at * with higher
precision. Of course this can only be determined by studying some
relevant test cases.

In any case I can propose one simple change to make the pattern more
specific. We can replace the n classification with the constraint

> ?.*Ox        connect
> ..OXX
> ..a..
> .....
> -----

oplay_attack(a,*,*)

/Gunnar