Re: computer-go: using move order information.

[Don Dailey <drd@xxxxxxxxxxxxxxxxxxxxx>]

Very good question!  I have  considered this question at great lengths
in  years past.  I  never  came to a firm   opinion on how useful this
information is.  I do know that using  it is full  of traps, but since
it is valid information, I can't help but feeling it could be useful.

There  are anecdotes  about humans  trying   to use  this information.
There have been  times  that I  used the  fact that  a master made  no
attempt to defend as an indication  that I didn't  really have a valid
attack, I did this  to save time on  my clock when I  was short of it.
It's an unreliable trick to use, trying to  cleverly use the opponents
knowledge to your  advantage.  Had I  reached the  same exact position
from a  different move order, I   may have tried  the attack,  and who
knows, I may have won!

Don




   From: "T.R. Christie" <trc23@xxxxxxxxxxxxxxxxx>

   my first mailing to the list.
   Interested to see what you think...

   As I understand it, most Go programs do not use
   any information about the order in which the stones
   on the board have been played, to make their move decision.
   To an extent this is understandable, because the best move
   in a given position will clearly be the same regardless of
   the order in which the stones were played.

   However, it is also true that the best move is more likely to
   be in the vicinity of recently played stones than not.  
   Also there are situations where move order seems to be
   useful in determining what to do.
   For example in this simple edge play black(x) should
   only play 1 if unconditionally prepared to follow white 2
   with black 3 (Assume x and o are unconditionally alive):

   xxxoooo
   ..312.. 
   -------

   but if we have reached, by some other sequence

   xxxoooo
   ...xo..
   -------

   black may well want to play somewhere else on the board,
   and not connect.  This seems to imply that move order information
   can be useful....(To provide some kind of continuity of logic?)

   So, can move order information be useful for computers
   playing Go, and if so, to what extent?