Re: computer-go: A problem with understanding lookahead

[Don Dailey <drd@xxxxxxxxxxxxxxxxxxxxx>]

   X-Sender: ddyer@xxxxxxxxxxxxxxxxx
   X-Mailer: QUALCOMM Windows Eudora Version 5.0.2
   Date: Tue, 23 Jan 2001 16:02:46 -0800
   From: Dave Dyer <ddyer@xxxxxxxxxxxxxxxxx>
   References: <20010123160847.E8740@xxxxxxxxxxxxxxxxx>
    <I1JHLB.A.i_E.Qegb6@xxxxxxxxxxxxxxxxx>
   Mime-Version: 1.0
   Sender: owner-computer-go@xxxxxxxxxxxxxxxxx
   Precedence: bulk
   Reply-To: computer-go@xxxxxxxxxxxxxxxxx
   Content-Type: text/plain; charset="us-ascii"
   Content-Length: 756

   At 02:59 PM 1/23/01, Heikki Levanto wrote:
   >Dave Dyer <ddyer@xxxxxxxxxxxxxxxxx> wrote:
   >> In go, there are no such metrics.
   >
   >Ok, idle speculation: What sort of metrics could be useful in go? Never
   >mindthe branching facor, just speculate on what we could possibly use for
   >determining if we like one position better than another?

   My point is precisely that there are no known metrics which are
   both useful and computable, and not even a theoretical
   framework exists that could lead to such metrics.

   Go has metrics "thickness" "influence" "alive" "dead" "heavy"
   "overconcentrated" etc, but they are all essentially uncomputable.
   Conversely, lots of computable things have been tried as proxies
   for the real metrics, and none has been shown to be much use.


All of the metrics in chess  are proxies for  the real things too.  In
chess,  the goal is  to checkmate  the king, being   up in material is
roughly corelated with the ability to do  this.  How much material you
have is computable but  doesn't have to  do  with being  in checkmate.
Being computable doesn't   mean much, 4+4  is computable  but probably
won't help my chess program.   

I'm telling you, we  have the same exact problems  in Go and in chess,
the  issue is just  clouded by the  fact that chess programs search so
deeply that people think intelligent evaluation is going on.

Don