I think your position is just too radical here.

Ultimately all science has to eventually be used for prediction or it is useless except for aesthetic purposes. However, I do sympathize with what (I think) your main point was before, that prediction is no measure of a theory if the "theory" is just curve-fitting (it is, of course, a measure of the utility of the curve or equation that the data was fit to). That is really just common sense, though, so you may have meant something else.