When do we have (and not have) evidence for a conspiracy theory (CT)? This talk shows how several tempting inferences when trying to evaluate the evidence for a CT have often unnoticed costs, which can be illustrated with Bayes' theorem. Canvassing these moves helps see why it is easy for most people to overestimate how much support a CT has. Conversely, it also shows CTs are an excellent case study for teaching people how to apply Bayes' theorem beyond thinking about base rates and odds.
(Attendance at the previous session on CTs is not required for this talk to make sense - the first is on concepts and priors, this is on evidence and updating).