Irish damselfly: And what do you think a theory is based upon?Well, yes, ultimately, there is only experience. Ideas come from something. Imagining the impossible is impossible. We only know that which exists.
However, in most fields of knowledge, what we know now is not greatly different from what we knew way back when. The Greeks and Romans knew that Earth is a sphere. They argued over whether it is the center of the universe. Aristarchos of Samos put the sun at the center with the Earth going around it. However, Archimedes performed close, careful experiments which proved that the Earth must be at the center, unless the universe is supposed to be impossibly large. That is just one example.
The Greeks (Aristotle, or one of his students, most likely) figured out the "vector" composition of forces, which lay intellectually dormant for about 2000 years. The ancients disavowed "irrational" numbers, and "negative" numbers. The medievals disavowed the "imaginary" square roots of negative numbers. These things all EXIST, but they lacked THEORIES to explain them. So, they were not well integrated into useful knowledge.
Thomas Kuhn called these major changes "paradigm shifts" in his Structure of Scientific Revolutions. The facts do not change. Our understanding of them does. Similarly, Ludwig vn Mises said that theory is more important than facts because theories explain facts, but, being a rationalist, he derived a perfectly logical theory -- praxeology -- that he claimed was "above" the mere facts. Von Mises pointed out that the socialists and capitalists do not argue over what the price of a commodity was at some time in some place. They have other disagreements, entirely.
David Kelley addressed the "problem of induction." He pointed out that we really do not repeat the same experiment over and over and over, building an inductive expectation of the outcome which is refined with further experiment, etc., etc. (That is the "fallacy of induction." It is part of the dumbing down of science education caused by the epistemology of public education.) Typically, one event is all we need to build a theory. We test theories, of course, and do so against, new, independent events, but, again, it does not take a long string of these, but only one, to test a theory.
It is a fallacy of our educational modalities that we expect scientists to laboriously measure and remeasure the light of stars passing near other stars and then abstract from those events a theory. In truth, Einstein saw the essence of the problem and proposed an explanation, which was tested with a single experiment.
Alternatively, Edison, the arch-empiricist experimented endlessly, with no theoretical basis for his works -- or so we believe. What great theories came from the Edison Labs? For all their experimenting -- hundreds of them working daily for decades -- what did they establish as truths?
Like the cones of geometry we have two fallacies. One says that tinkerers discover facts, which engineers generalize into inventions, which scientists explain with theories. The reflexive fallacy is that scientists develop theories which engineers turn into practical applications which technicians maintain and improve.