Re: computer-go: FPGA

[Don Dailey <drd@xxxxxxxxxxxxxxxxxxxxx>]

   From: Dave Dyer <ddyer@xxxxxxxxxxxxxxxxx>
   Cc: Mike Gherrity <gherrity@xxxxxxxxxxxxxxxxx>

   At 03:56 PM 8/30/00, Mike Gherrity wrote:
   >Hi Don,
   >
   >> Dave, I  am interested in why a  computer would get consistantly worse
   >> results   with more computing  power  and   why  you believe it's  not
   >> possible to fix.

   I didn't claim that more cpu wouln't improve the search, only that
   you couldn't guarantee that it would.  Since your evaliation is imperfect,
   searching deeper or wider will show you different things, and you might
   temporarily perfer an inferior move.


Yes, I understand that.  In chess sometimes a  deeper search prefers a
weaker  move, but  on the average,  a  one ply deeper search  will win
about 3 to 1 and this seems pretty much  guaranteed, at least with any
reasonable evaluation  function.     But I  can   imagine writing   an
evaluation function that makes it play worse with depth (reversing the
sign of the score for instance!)

Don