A Google with search term: polynomial regression "r squared" ; gave 37,000
hits
This one look useful for your purpose
http://www.stat.wisc.edu/courses/st333-larget/poly.pdf
best wishes

=LINEST(ydata,xdata^{1,2,3},const,TRUE) returns R^2 in the 3rd row, 1st
column.  You can use the INDEX() function to return a value from a specific
location of an array.

If you want to calculate it from first principles, then
R^2 = SSR/SSE

While not the numerically best approach (which is not closed form),
SSR = SUMSQ(predicteds)-n*AVERAGE(ydata)^2
SSE = SUMSQ(ydata)-SUMSQ(predicteds)

If you are not fitting a constant term, then do not subtract the correction
term
n*AVERAGE(ydata)^2

Jerry