Predictability and Understanding

I've drawn a schematic that classifies how I see various physical theories. The y axis shows how successful the theory is at predicting the real world, or its testability. For example quantum mechanics (QED) can be tested and is very accurate, whereas Lovelock's Gaia Theory is not very predictive or testable, yet, and string theory is not at all. On the x axis I have Understanding, which is supposed to be a rough measure of whether the theory comes with a physical mechanism that explains intuitively how it works. So, thermodynamics is well explained as being due to the statistical motion of atoms, but the onset of turbulence in fluid dynamics is not at all understood:

The top left box is for empirical theories which predict the world quickly, but without the need for an understanding. They usually have adjustable parameters. Many models have been born in this box and end up migrating to the right into the physics box when they are understood better, eg: quantum mechanics originated from Planck's empirical formula, with the adjustable parameter being Planck's constant.

On the top right is the physics box, which mixes predictability with some understanding. The ideal physics is in here. Ideal physics can predict nature, but also includes an intuitive mechanism which helps us to extrapolate the ideas to other parts of nature. I've listed thermodynamics and special relativity very high on both axes since the mechanisms for both are well understood and they both make testable predictions that fit nature.

On the bottom right is the box for philosophy which is all about understanding, without necessarily a need for directly predicting details of the real world. Some beautiful ideas that end up as part of theories start off here, like Mach's idea about observables which helped to produce special relativity.

I'm not sure what to call the bottom left quadrant, but here the theories are neither well understand, nor do they make any testable predictions about nature. This is the worst kind of model: untestable. The kind that Pauli would have called: "Not even wrong". Most of contemporary theoretical physics is here. This is probably the case in every epoch: the only theories we remember from the past are the better ones that survive. These tend to be born in the fertile empirical box and migrate slowly to the right, sometimes with a dash of insight from the philosophical box.