Re: computer-go: perfect play : a new fractal algorithm

[Serge Boisse <boisse@xxxxxxxxxxxx>]

Joan Pons Semelis wrote:
> 
> The idea looks workable for strategic decisions about important
> areas to play, but tactically go is not scalable. Just take a 3-3
> invasion: you still need 2 eyes to survive, even if they are small,
> and only 1, even if it's as big as a rabbit 6, makes a dead shape.
> Unless you change also L&D rules, using for example immortal stones,
> a go variant pretty interesting.
> 
> Joan

That's right.
This is why a local search is nessary at each subscale, to account for this
tactical complexities. But the number of move to explore for these local
seraches still can be reduced by the "genius advices" uploaded from the
immediate subboard. So the search is tractable, and may be 15 or 20-ply, which
is enough for most tactical needs.
So i don't think that your (good) remark is killing my idea.

Also the fractal stuff might be only a part of a whole programm, which might use
both fractal "advices" and more classical ideas (patterns, neural net...) to
select the bestmove at a given time.
-- 
                    ____________ 
                   / Linux now !\     ("`-/")_.-'"``-._    _
          ,,,    O \____________/   o  @ @ `; -._    )-;-,_))
         /'^'\ o°    / catch it ?\O°  (v_,)'  _  )`-.\  `---'
        ( O O )      \___________/  __.- _..-_/ / ((.'
 +---oOOo-(_)-oOOo-----------------((,.-'---((,/---------+
 |  Serge Boisse                                         |
 |  SERVICE TECHNIQUE DE LA NAVIGATION AERIENNE (STNA)   |
 |  ODS FRANCE project, http://www.stna.dgac.fr/phidias  |
 |  tel: (0)562 14 5731      mailto:boisse@xxxxxxxxxxxxxxxxx  |
 |  homepages: http://www.multimania.com/boisse          |
 |  and:       http://www.multimania.com/unitedplanet    |
 +-----------Oooo----------------------------------------+ 
     oooO   (   )
     (   )   ) /
      \ (   (_/
       \_)