See, when I was in grad school I once had to calculate an agreement metric from a bunch of labels on a corpus. No problem I said, the math is easy, I can write a script in an hour or so. Fam that mfing script took me two freaking days bc there were always some little bugs or weird edge cases I hadn't thought of. So the deal I made with myself was: I would use Matlab or a stats library or something like that, BUT I would make sure that I understood the math beforehand.

But for whatever reason, I never had to calculate a standard deviation. Thinking about it, someone else might have done that for papers I was co-author on, though.