A scientific theory, as opposed to the colloquial use of the word, is more than just a guess, nor is it, as Isaac Asimov once quipped, something you came up with while drunk.
It is a constellation of ideas that weaves facts together into a single overarching description and detailed explanation for a given set of phenomena.
Any scientific theory is provisional, never proven with complete certitude, but it is important to distinguish a theory from the facts it describes.
The scientific use of a theory gives no indication of its actual degree of certainty, though a given set of such ideas may be so well established by repeated testing as to be demonstrated as confirmed beyond all rational doubt, such as the theories of genetic inheritance, general and special relativity, quantum mechanics, number theory in mathematics, music theory in music, and stress theory in engineering, the germ theory of disease, atomic theory, heliocentric theory, round Earth theory, plate tectonics theory, and of course, evolution.
Unfortunately not everyone’s doubt is rational, and various sorts of science deniers are given to labeling any set of scientific ideas they have a bug up their proverbial butts about as ‘just a theory, not a fact,’ playing on the general misunderstanding of both the words ‘theory,’ and ‘fact.’
A theory is not a hypothesis, as the latter is merely a component of a theory, a specific set of predictions within its overall framework, what one should expect to see, and not see, if that part of the theory is to be confirmed or falsified.
A fact in science is never absolute, due to the provisional nature of the scientific enterprise, for it can never be known with complete surety that some data which may disconfirm that fact will never appear at any given point in the future.
Theories are never ‘promoted’ to laws, as they are two different sorts of beasts — A law merely defines things, while a theory describes and explains how they work.
Theories usually start as models, which propose hypotheses for testing by experiment or other observation, such as the comparative method in the mostly historical sciences, such as geology, cliodynamics, cosmology, astronomy, paleontology, and archaeology to name a few.
No, you do not have to do experiments in a lab to do science. Otherwise, no crime that has ever been committed could be solved using evidence left at the scene it occurred at, and detectives would be permanently out of work as a profession. Science is not just for the nerdy guys (and gals…) in labcoats and pocket-protectors.
Most science today is done as a community effort, evolving over time, and all involved in a study contribute something to the overall theory being investigated; the idea of the lone gentleman researcher working in his basement lab, the sole progenitor of his ideas, is a quaint notion, however popular it may be in fiction.
Even broader than a theory is an overarching concept called a paradigm, often composed of many theories. M-theory in cosmology would be a good example of a paradigm, as a candidate for a ‘theory of everything’ composed of many subsets that individually describe and explain some aspect of reality.
The term was coined by philosopher of science Thomas Kuhn and he used it in an early attempt to describe the internal process by which science changes, though it is now more often used in the sciences to refer to a conceptual tool, as a mode of thinking or general working approach to theories and frameworks of theories.
Further, a theory is never supported by only one piece of evidence, but through multiple, often thousands, even millions, or more independent lines of data spread throughout multiple research disciplines, which is why the demand science deniers make of, “show me just one piece of evidence that proves the theory true,” is nothing more than an empty rhetorical stunt, and an unfair shifting of the burden of proof.
Ultimately, any theory can never be proven to be timelessly, absolutely true by a finite data set, though depending on the nature of the theory, it sometimes only takes one reliable observation to falsify one or more attendant hypotheses.