This way you can even test a hypothesis you don't even understand. Just get a prediction out of someone who does and see if it checks out without being otherwise obvious. Again, obvious beforehand, not afterwards. There are a couple of heavy tomes you can read about it, a couple of common tricks to side step the test (ad hoc hypotheses) but in principle that's it.

For example, you test the theory or relativity every time you use GPS. It usually works.

Everything else is a matter of taste. Do you like consistency? Logical structure? Elegance? Authority? Peer review? You can make up whatever test you like. There's bayesian approach, you can try to consciously correct for your biases... but it's hard to beat the scientific method.