But what good is a theory which doesn't make testable predictions?
We don't regard epicycles as a valid theory any more because it could make no predictions. With sufficient fiddling, epicycles accounted reasonably well for the apparent motions of the planets. But it said nothing (and could say nothing) about the motion of a new planet, if one were to be discovered. You could only make observations and set out to find some arrangement of "gears" which matched. It would be an after-the-fact "fudge".
Whereas Kepler's laws had predictive power, They set limits on the possible courses on any new worlds.
Even if string theory is correct, we have 1e500 variations and, at present, no way of telling which one we live in. The masses of particles and the strengths of forces have to be "plugged in".
I'm aware this situation might change. Mathematical consistency might rule out all but one version. Or someone might come up with a clever experiment which didn't require an accelerator the diameter of the galaxy. So far, all we have to show for it is;
1) some very elegant math, and
2) a firm prediction that a zero-spin particle must exist for everything to work out properly, so gravity is automatically included.