10 comments

[ 4.6 ms ] story [ 30.8 ms ] thread
> The applicants, if all seen together, can be ranked from best to worst unambiguously.

Sounds to me like the actual objective is to hire the hottest-looking secretary.

Well, it was formulated in the 50s.
This problem is the underpinning of lots of real-life problems. Basically any search where revisiting a previously evaluated choice is expensive.
This can be made much more efficient if you just take all of the applications and throw half of them away.

This will ensure that all the remaining applicants are luckier than the average applicant.

I wonder what happens when there's lots of noise in the estimation of quality or fit (or whatever underlies the preference ranking). I imagine this has been done?
I learned about this in the book Algorithms to Live By. They talk about it in the context of dating; How do you know if you've met the right one? IIRC you have to go on x amount of dates knowing you won't commit to this person.

The book has an interesting way of looking at life, recommend.

>> is simple and selects the single best candidate about 37% of the time

Not a mathematician but 1 / e = 0.367, so is this just a way to ensure a 37% percentile of accuracy across normal distribution? Like Anchorman Sex Panther, "37% of the time it works every time"?