If I make a prediction that a certain but yet unspecified person has the ability to run 100 meters in under twelve seconds, which is near Olympic standards for time, two things should be apparent: There are observations that wouldn’t be expected by this theory, like the person simply trying and failing to accomplish this feat, and that the prior probability that any given person can achieve that is quite low.
However, our confidence in this claim should rise, somewhat slightly, upon finding out they are in their late twenties. The probability should again rise if more information revealed this person ran track in high school and would increase even more if you were to receive information this person was a collegiate sprinter. At any stage in this early process unfavorable evidence, like revealing the person is sixty instead of in their twenties, would lower the probability that they could achieve such a physical feat. Ultimately we can settle this with great precision by timing this person run a 100 meter dash on several occasions.
What’s so mundane about this is that there is a conceivable set of evidence which would increase the probability they could beat that time, up to and including physically testing them to see if they can achieve it and a similar set of data my theory would not predict which would decrease my confidence in this claim. This is how predictions work, however, this is nothing like the claims people make about alleged supernatural or pseudoscientific abilities. Today supernatural and pseudoscientific proponents are largely united in either making no predictions whatsoever or making testable predictions that are not met but then denying that the overwhelming observed evidence which their theory doesn’t predict should lower our confidence in their claims. In other words, they are either making unfalsifiable claims or making claims that have been tested and shown to be false.
Instead of an acknowledging there is observable data which would count for and against these claims what one finds instead is that many simply assert their ideas can’t be tested as if that’s makes their idea plausible. To any such claim the simple question of “What does your theory not predict?” is perhaps the most revealing of all possible lines of inquiry. When faced with this believers have to acknowledge that their theory predicts everything and is therefore an unfalsifiable claim, admit that their claim can be shown to be demonstrably wrong (which usually means acknowledging it already has been), or by exiting the present conversation. Naturally, as useful as this question is, this means if you are going to get into the habit of asking “What does your theory not predict?” of believers, you’d better get used to people fleeing in apparent horror.