Re: [gnugo-devel] trevor_1_17.7

[Gunnar Farneback]

Trevor wrote:
> http://www.public32.com/games/go/trevor_1_17.7
>  - removed small value lunch attacks & defends from owl.c
>  - lots of tuning

>From the patch:
> +Pattern D721
> +#tm New Pattern (3.1.17)
> +
> +.X?  
> +X.X
> +.*X
> +

This pattern had better be an Axxx pattern.

>  Pattern D1310
>  #tm modified (3.1.15) (see trevora:400)
> +#tm added b constraint (3.1.17)
>  
>  ?ooo
>  ?ooo    try to escape
>  X*oo
>  ?OX?
>  
> -:8,-,value(80)
> +:8,-b,value(80)

Does the b classification really make any difference here?

>  Pattern EB713
> +# tm modified (3.1.17)
>  
>  ?O..          block
>  X*..
>  ----
>  
> -:8,OXb,followup(3)
> +:8,OXb,reverse_followup(3)
>  
>  ?Ob.
> -X*a.
> +C*a.
>  ----
>  
> +;!owl_threatens(*,C) &&
>  ;!xplay_attack_either(*,a,b,*,b)
> +

I agree that reverse_followup is more appropriate than followup, but
why should this not be awarded if * is an owl threat against C?

> -Pattern Conn106
> -
> -??.?        indirect connection
> -*O.O
> -....
> -
> -:8,C
> -
> -??e?
> -*OaO
> -.dbc
> -
> -;oplay_attack(*,a,b,e,c,d,d)
> +#tm removed (3.1.17)
> +#  this is a very strange way to connect.
> +#  see, for example, buzco:8

That's beside the point. Sometimes this connection is the only way to
connect while simultaneously managing to do something else, e.g.
tactically capturing some stones. Thus the move should have a
connection move reason. If it's an inappropriate way to play in some
context, this should be taken care of by other means, typically shape
values, replace patterns or antisuji patterns.

/Gunnar