Candidates arrive in random order and must be hired or rejected immediately. Move the cutoff: the first k are only observed, then the first one better than all before is hired.
The science
This is the classical secretary problem. With n candidates in random order and a rule that rejects the first k and then takes the first candidate better than all previous ones, the probability of hiring the single best is P(k) = (k/n)·Σ_{i=k+1..n} 1/(i−1) (and 1/n for k = 0). For large n this becomes −x ln x with x = k/n, maximal at x = 1/e ≈ 0.368 with P = 1/e ≈ 0.368.
The Monte Carlo dots on the graph test every cutoff at once on each random permutation, and the page reports a z-score against the exact value as the independent check.
Simplifications: candidates can be ranked without ties, you only care about hiring the very best, decisions are irrevocable, no one declines, and the pool size is known. Real recruitment has noisy scores, offers that are declined and partial credit for a good (not best) hire, so treat this as a model of one trade-off between exploring and committing.