Re: [gnugo-devel] bad fuseki

[SP lee]

> GG.3.5.3 falls into panic after black plays 27 L15
> it adds tons of moves to secure the top edge and capture the black stone.
> Move 36 M16 is the worse .
>
> I wondered about the move reasons for 36 M16
> I just found Pattern CD107
>
> Then i saw that Pattern CD107 does not match any more for 3.5.4
> (pattern modified) , but 3.5.4 still like 36 M16 very much .
>
> M16: 21.55 - change in territory.
>
> notice that 3.3.16 played M16 at move 28 and
> add negative territorry  value for 36 M16
>
> Also , what about
> ; !dead(d) && !weak(A) && oplay_attack(*,a,b,c,c)
> which seems to be more general than  Evan's Pattern CD107 modification .
> (d being A and A becoming the X stone as the conventions need) ?
> md.
>

Actually I'm thinking about adding the constraints !oarea(*) to many
patterns including this, especially some escape patterns. Maybe we could add
a new classification to prevent inefficient moves?


SP Lee