I found one of the most fun CS-augmented stats courses gradually built a stats library bottom-up over the time period of the course using simple constructs to solve lab assignments.

These are elementary concepts yet the interviews are regarded as very hard, so there must be more to it...

Theres a difference between knowledge and intelligence. You might know these concepts, but could you apply them in novel situations under pressure in an interview? Could you maybe link disparate ideas you know together to solve a completely unknown problem?

For example, by 11th or 12th grade you should, in theory, have all the necessary tools to get a perfect score on the AMC10/12. However, unsurprisingly, very few people actually do.

Thats why these interviews are hard, they don't actually require anything that advanced. I hate to use "IQ" as a metric but these types of questions/interviews tend to screen more for raw skill than they do for knowledge. They require you to actually understand and think about the problem (and execute within a short timeframe), which, as others have already mentioned, is nontrivial for the vast majority of people.

Source: was at a similar trading firm and known several people at other firms (incl Js).

This is mostly to get high schoolers interested in probability and to make the work done at Jane Street seem more accessible

Eh, most people I interview couldn't write an iterative DFS. I think the bar for a hard interview is pretty low.

agreed