You're discounting the possibility that "fringe" medical advice would have theoretical evidence behind it, as opposed to empirical evidence.

Are you talking about a theory rooted in solid biological and medical science or some alternative newagey theory like chakra points? Even within the realm of scientific theories, the ability to predict the actual effectiveness and safety of a therapy is generally limited: 92% of drugs that pass in vitro tests fail animal or human trials, and these are only the drugs that have already passed the computer-based design phase.

You're also not doing an expected utility calculation. What are the costs and benefits of following this particular bit of "fringe" medical advice? In my case, the cost was: I spent a few hours reading and massaging my arm, and discovered a knot in my bicep such that when I massaged it, my wrist pain was replicated.

As far as I know, there are a number of conditions, notably inflammation, that are actually made worse by massage.

The fact that someone is passing it off as medical advice makes the probability of it being useful medical advice way, way higher than the probability that some random string of characters is useful medical advice.

That's an irrelevant comparison, since nobody suggested to sample random strings for medical advice. In any case, an alleged medical advice also has a probability of being actually harmful way way higher than that of a random string.