If it's a silver bullet, the randomized experiments must not have included any werewolves among the subjects: see discussion in "Preventing future offending of

If it's a silver bullet, the randomized experiments must not have included any werewolves among the subjects: see discussion in "Preventing future offending of delinquents and offenders: what have we learned from experiments and meta-analyses?", MacKenzie & Farrington 2015 https://www.gwern.net/docs/sociology/2015-mackenzie.pdf https://www.gwern.net/docs/sociology/2015-mackenzie.pdf (Note that it is an old debate in criminal justice based on the randomized experiments' dismal results: "does anything work?" The iron laws of evaluation apply as much there as in the rest of sociology...)