pass on the first 37%, then hire the next one who beats your best so far. this finds the absolute best candidate about 37% of the time. if all you can see is how each candidate ranks against the ones before, no strategy does better. here you also saw the scores, which is extra information: a rule that uses them can beat 37%.

why skip anyone? because a score means nothing until you know what normal looks like. the skipped candidates are your benchmark: they're how you recognise "good" when it walks in. skip too few and your benchmark is weak, so you hire the first decent one. skip too many and the best has probably already left.

why 37%? say you skip a fraction r of the line. you win exactly when the best candidate comes after the skip and nobody between the skip and them beat your benchmark. if the best sits at position x (as a fraction of the way through), that second part means the best of everyone before x was in the skipped stretch, which happens with chance r/x. averaging r/x over every x from r to 1 gives -r·ln(r). that peaks at r = 1/e ≈ 0.37, where e ≈ 2.718 is the base of the natural log. and the win chance at that peak is also 1/e: the fraction to skip and the odds of success are the same number.