The concept of 'living in the matrix' is only meaningful if there is some difference with not living in the matrix. Thus, there should be something, at least in principle, that we can observe. And based on this observation (or lack thereof) we can determine probabilities.

Regarding the 50%, suppose (for the sake of argument) that there really is some property of the universe that is either true or false, but we have no evidence whatsoever in either direction. Then the probability is simply unknown, not 50/50. Of course, as a Bayesian prior starting point, you can use 50%, but that is only a rule-of-thumb, and shouldn't import all the connotations of a "50%" in cases where we do have facts.