the point equivalence classes are easy enough to define.  points are
either in a size-1 class (the center point, can be ignored), a size-4
class (axes, both vert/horiz. and diagonal), or a size-8 class (all other
points on the board).

for each tuple of points in the equivalence class, normalize whatever
measure you're using for the importance of those points (varies according
to what you've implemented, of course) to sum to 1.  then calculate the
mean and standard deviation of the values of the points in the tuple.

sum the standard deviation over all of your tuples, and you should have a
reasonable full-board measure of the symmetry of your algorithm that can
be compared with other people's measures.

not a rigorous statistical measure of anything by any means, but it ought
to be fair enough for what you're suggesting.

i think that a more rigorous (or just overblown and more painful to calculate)
measure would be the sum over tuples of [width of 95% confidence
interval using "tuple-sized-1" degrees-of-freedom student's t distribution] 
after
normalizing each tuple to sum to 1.

s.




 
____________________________________________________________________________________
Looking for earth-friendly autos? 
Browse Top Cars by "Green Rating" at Yahoo! Autos' Green Center.
http://autos.yahoo.com/green_center/
_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/

Reply via email to