On Thu, 2008-10-09 at 14:01 -0700, terry mcintyre wrote:
> There's a huge difference ( perhaps hard for a monte carlo program to
> grok ) between "Against a skillful oponent, my group is provably dead
> no matter how I try to defend it" and "My group is alive in a
> statistical sense, since there is only one statistically improbable
> line of play which my opponent could use to kill this group."

If the computer thinks it's winning with a 90% score but loses,  was the
statistics right or wrong?   Statistics are almost meaningless with one
data sample point.   You need to look at 10000 positions where it said
90% and see if it usually wins or loses if you want to evaluate their
judgment.  

The opponent model a computer uses (the play-outs) is not based on an
actual opponent (indeed, this is not possible because the computer may
play a different opponent each time) so the scores returned should not
be viewed as an actual winning probability.   It's simply a
self-consistent evaluation function which is imperfect, kind of like us
humans.



- Don




> 
> That "statistically improbable" line just happens to be the one which
> a skillful opponent will latch on to, oddly enough.
> 

Attachment: signature.asc
Description: This is a digitally signed message part

_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/

Reply via email to