In this case I have assumed that your ability to estimate the promise of therapies is actually worse than random. In this case you tend to guess that the therapies which are actually the least promising are the most promising. So in this case despite a reasonable cutoff, you pick the therapies least likely to benefit you.
Is such a thing actually possible? I think that it is. For instance, if you took all claims by proponents of alternative therapies at face value without considering the quality of the evidence, you would be attracted to the therapies with the most outrageous claims - and I would not be at all surprised if these turn out on average to be the less effective therapies!