Re: computer-go: Live or Die

[Vincent Diepeveen <diep@xxxxxxxxx>]

At 11:31 PM 12/9/99 +0100, you wrote:
>
>Tim Klinger schrieb:
>> 
>> On Mon, 6 Dec 1999, Jeff Massung wrote:
>> 
>> > Could someone give me some examples of eyes that are more than 1 space in
>> > size?
>> >
>> > Thanks.
>> > Jeff
>> 
>> x x x
>> x . x
>> x . x
>
>It is probably always bad for X to play b2, but it is
>not always a bad play for O, as the example below shows.
>
>  . . . . . . . . . .  
>  . o o o o x x x . .  
>  . o x x x o o x . .
>  . o x . x . o x . x
>  . o x . x o o o x .   Lower edge, O to play
>  . o x x x o . o x .   (outer stones alive)

>It is hard to indicate "bad moves" from a pur local view of
>e.g. 3x3 or 5x5 regions extracted around a free point from 
>a full board position.
>
>Candidate move at ? good (or even necessary) or bad, 
>local 5x5 region of example:
  
>    o o o o x  
>    o x x x o 
>    o x ? x . 
>    o x . x o   
>    o x x x o


    o o o o x  
    o x x x o 
    o x ? x Y
    o x . x o   
    o x x x o

interesting problem, but isn't it such that the open space marked with Y
is part of the precondition of the structure:

xxx
x.x
x.x
xxx

Which might say:
  if( all open spaces around structure can be filled )
    then EVALUATE structure;

So the precondition is simply not met, which means
the 5x5 subboard can never apply here!

>There are already a lot 5x5 regions around a free point.
>My (almost) random Go playing program marks ~10..20 points 
>as "bad" when comparing 5x5 subboards agains a database 
>of 5x5 patterns. (of course only near the end of a game, with 
>already a lot of unnecessary moves -by human standards- made).

>Hans
>
>
Vincent Diepeveen
diep@xxxxxxxxxxxxxxxxx

---
... en verder ben ik van ben ik van mening dat Dap het gesticht in moet