you should also consider the amount of cases where people said the same "he's kind of a weirdo", but that person did not go on to do something horrible. And also the amount of cases where people did not say it, and yet the person did something horrible. All three are necessary to calculating the strength of the evidence "people say he's kind of a weirdo" in favor of the hypothesis "he will do something horrible"

Well, yeah, I agree, but I wasn’t trying to do that. At least I don’t think I was, and if it’s an implied assumption in what I said I don’t see it.

My original comment just said that I suspect many of the “I had no suspicion” after-crime statements are false (consciously or not), and was based mostly on how I suspect people’s brains might react, not on the rates of horrible acts.

My second comment I think said the same thing your quote above does, except adding that I also suspect a certain ordering of rates. But as I said in my first comment, I don’t have the actual rates and I believe they’re hard to obtain, so it’s just a suspicion.

After all, I don't expect a random stranger on the street to suddenly commit a horrible deed; why should I expect it more of an acquaintance unless there are concrete warning signs, which would make me say "I always knew he was up to no good".

The true opposite of "I always knew..." would on this view be like Harry's reaction to Hermione confessing attempted murder: "I don't believe she did it, it's a priori so improbable there must be another explanation or very special circumstances".

That’s true. I guess there are just very few people with this kind of reasoning (à la the Wizengamot); once they heard it happened, most probably take it for granted it was so, and they have only the “knew it all the time”/“didn’t see it coming” alternatives.

After all, I don't expect a random stranger on the street to suddenly commit a horrible deed; why should I expect it more of an acquaintance unless there are concrete warning signs, which would make me say "I always knew he was up to no good".

For a random stranger you have only the base rate to go on, you’ve got no other evidence. (Though for specific strangers you might have stuff like “he looks like a mobster” or “I’m in a dangerous neighborhood”, or maybe “he’s black and wears a hoodie”, which are a bit different as signs go.)

My claim is not quite that “weird people murder more often”. Instead, I suspect that “of the people who murder, a big majority were not stable/calm enough before and did give signs before”, and many if not most of the cases of people claiming there were no signs are because they just forget or ignore those signs.

(The two sentences are different if it so happens that there are very few people that give no signs, enough so that the fraction of them who do horrible things is less than the fraction of those who did give signs. Which I believe unlikely but not quite impossible.)