Can you have a look at this?
The [/w/index.php?title=User_talk:Armondikov&offset=20130318140948&lqt_mustshow=14553 highlighted comment] was created in this revision.
It's mostly right based on Yudkowskian rationalism, which is a fairly robust approach to thinking about hypothesis testing but unfortunately suffers from too much subjectivity when trying to assign an actual, i.e., numerical, prior probability to things. Indeed, my snarkastic view of Bayes theorem is that you just fiddle with P(B) until it gives you the answer you want.
Overall, it's not the best wording of it, I must say. But the subject is worth including in the article on evidence. The example is okay. You expect to see Christians exist in the world if God exists and you expect to see Christians in the world if God doesn't exist but religion is a social phenomenon. The question is which is more likely as does this "evidence" actually tell you the difference.