Excel computes R2 (r squared) incorrectly for the case where a quadratic fit is obtained through a set of data scattered about a parabola.
The correct value of R squared for this case should alway be near zero, due to the definition of R squared.
This raises an important question: how does Excel calculate R squared in the first place?
This result was obtained using the curve fitting feature of a graph in Excel Office 12.