Re: computer-go: Good Play (was FPGA)

[Don Dailey <drd@xxxxxxxxxxxxxxxxxxxxx>]

   Date: Sat, 02 Sep 2000 08:27:29 -0500
   From: Matt Gokey <mgokey@xxxxxxxxxxxxxxxxx>
   X-Mailer: Mozilla 4.7 [en]C-CCK-MCD NSCPCD47  (Win98; I)
   X-Accept-Language: en
   MIME-Version: 1.0
   References: <200008310756.AAA17426@xxxxxxxxxxxxxxxxx>
   Content-Transfer-Encoding: 7bit
   Sender: owner-computer-go@xxxxxxxxxxxxxxxxx
   Precedence: bulk
   Reply-To: computer-go@xxxxxxxxxxxxxxxxx
   Content-Type: text/plain; charset=us-ascii
   Content-Length: 1756

   Mike Gherrity wrote:

   >> The problem is that people don't do a brute-force, iterative deepening
   >> search.  How do they play so well?  It would be nice to figure out the
   >> answer to this question using a game like Chess or Go.  However, if it
   >> can't beat the brute-force, iterative deepening search machine, people
   >> aren't interested.

   How do they (people) play so well?  Hmmm, good question which got me
   thinking about an assumption we are making.  Computer Chess really is
   simpler than Computer Go for many reasons.  I don't think anyone here
   would argue with that statement.  But is Chess a simpler game than Go? 
   If so, how much harder is Go than Chess? It is more than a question of
   degree of difficulty, since the two games are very different perhaps
   requiring different skills, and so might be like comparing apples and
   oranges, but just how hard is Go?  Let's assume it is around a couple of
   orders of magnitude more difficult for the purposes of this discussion.

   Given that thought, the question that popped into my head was How do we
   know people (even at the professional Dan level) are playing the game so
   well? Could it be that even professionals would be devastated by a
   perfect/near perfect Go-playing entity?  If the game is so much more
   difficult than chess, are Go players much more intellectually capable
   than Chess players?

   Taking the opposite conclusion, perhaps Go really ISN'T much harder than
   chess, it is just different than Chess and due to its unique
   characteristics is not amenable to the brute force, iterative deepening
   search machine.

   Questions:
   Is Go inherently more difficult than Chess?
   If so, is there a way to quantify the degree of difficulty?
   Do expert Go players really play near perfectly?

   Thoughts?

   Matt


I have always believed  that people don't play chess  OR go very well.
But we suffer  from our own point of  view.  The  strongest Go masters
are so impressive to us that we don't  realize how "badly" they really
play.  Of course  I am only  guessing because it's  not  easy to prove
this.

It  has been demonstrated by using  Chess endgame databases that there
are   certain endgames that seem   to be virtually incomprehensible to
humans.  Now an  endgame database plays PERFECT  chess,  for its given
domain.   When strong masters used  the databases to  train with, they
still didn't get  it and could not see  or find the patterns (why does
this move win and this other move doesn't?)  Perhaps if they had spent
months or years on this they  may have made progress, but nevertheless
this result really clearly shows that we have a long way to go.

There are  other issues however.  I  believe the style  of play  is an
important  factor.  If GOD  decided   to play Chess against  Kasparov,
would he simply play  perfect Chess, or would he  add tricky moves  to
confound Kasparov, taking further advantage of what he knows about his
mental weaknesses?  If the opening position is a draw, then offering a
draw  on the first   move is a  "perfect" move  in the game  theoretic
sense, or quickly setting up  a 3 move  repetition.  It might also  be
that  rapid exchanges  that  quickly  lead to   a draw  might even  be
negotiated  by  Kasparov.  Even strong  players try   to make things a
little more complicated,  especially when they want  a win and a  draw
will not do.

So if I had a computer that was so fast it could see to the end of the
game, I  would spend some effort on  obfuscating the play.  There must
be some bluff  involved and some assumptions  about what makes it more
difficult for the opponent to get the maximum results!


Don