There are tests that don't use a standard deviation of 15. Feynman might have done, for example, the Cattell Culture Fair Intelligence Test[1], which has a standard deviation of 24.
Ability isn't well distributed to where it can have a maximal benefit in a given field. It's often said that people as smart as Einstein (and Feynman) are currently living and dying on subsistence living farms.
What hasn't been said, but is equally true, is that people as capable in physics as Einstein as currently working as waitresses in No Where, Idahoe. Or they're slaving away as bit-actors in Hollywood because that's what they love.
Or they're sitting behind a computer, typing away on HN, never to know that they're potentially amazing theoretical physicists simply because their 7th grade teacher turned them off to the idea of physics in general so they studied computer science instead.
Until there's some kind of test that tells you your aptitudes at very subject, you'll always have people pick suboptimal paths to excellency and something other than what they absolutely theoretically could have been best at.
[1]: https://en.wikipedia.org/wiki/Cattell_Culture_Fair_III