Re: computer-go: minimax and go

[David Fotland <fotland@xxxxxxxxxxxxxxx>]

1) much slower full board evaluation (20/sec vs 200,000/sec for chess)
2) much deeper searching required (30 or 50 ply? vs about 10 ply for master
play for chess)
      if you have specialized ladder or low liberty reading code, maybe
full board only needs
      to be 20-30 ply for reading out difficult semeai and life/death
fights on full board
      involving several groups.
3) much higher branching factor (100? vs 25 or so for chess)

But on the other hand, move sorting is probably easier than chess.  With
alpha-beta searching
about sqrt of branching factor, go search time is 10^25/20 = 10^24 secs and
chess search time is 5^10/200000 = 48 secs.

It will be a very long time before computer hardware advances erase 22
orders of magnitude
in search time :)

-David Fotland

At 12:57 PM 11/7/00 -0500, you wrote:
>
>Hi!
>
>I need to know (succintly) what are the arguments that a brute-force
>minimax approach to Go won't work assuming increased computer power 
>just like it worked for Chess? 
>
>I know it has probably been discussed here but the search capability
>on the archives is still not working. 
>
>Thanks,
>
>                                            myriam
>
>