Depends on the curve though. You've got to worry about it being both smooth and differentiable (I think) - books have been written on this <img border="0" title="" alt="[Smile]" src="smile.gif" /> . .. .For example, if you're trying to work with tan(theta), it breaks down at theta = pi/2 radians as the function is discontinuous there. . .. .If you're really keen on this then I'd look up a suitable method in <a href="http://www.ulib.org/webRoot/Books/Numerical_Recipes/bookfpdf.html" target="_blank">Numerical Recipes</a> and then try to implement this using Excel - if that's possible. . .. .Another possible approach is to fit something differentiable to your data and then differentiate it yourself, which will give you an analytic expression for the gradient at any point. For example, say your data is well fit by a polynomial y(x) = x**3 - 2x + 1, this can be differentiated trivially to give the gradient y'(x).. .. .It's hard to know without the function <img border="0" title="" alt="[Smile]" src="smile.gif" />