Are unforseen things determinable?

Using an example, in this thread, Cecil is explaining about the phenomenon of cockroaches dying on thier backs, when he says:

(Emphasis mine)

Let’s say that we have an event like this in which the components are unknown (or not thoroughly studied). How is it that we can determine (using the known principles of science) the event (in it’s entirety) and still be 95% sure that it will occur (in the way that we have mentioned)?

I guess what I’m trying to ask is, do we ever know that theory alone can account for an occurence? Or do things always have to be tested (via experimentation) in order to get an answer (like when trying to work out constants or distances from uncharted territories like space)?

Note to Mods: I’m not entirely sure whether or not this should be in GD. If you think a move is necessary, please make it…

Alas, the answer to your query is known only to Schrodinger’s Cat, who at this moment is in a box(or not).

“do we ever know that theory alone can account for an occurence?”
Not with certainty. But if we modify our theory every time we find that its predictions and external reality are at odds, we can ensure that the theory will correctly predict the outcome an ever increasing number of events.