How this instrument works
This instrument runs the math behind the 'secretary problem,' a classic result in optimal-stopping theory: if you must evaluate candidates one at a time, decide on each immediately, and can never go back to one you've passed on, the strategy that maximizes your odds of picking the single best candidate is to reject a fixed opening fraction outright, then accept the next candidate who outranks every one you've seen so far.
That opening fraction is 1/e, roughly 36.8%, where e is Euler's number (about 2.71828). For n total candidates, the rule says: reject round(n/e) of them as a benchmark-setting phase, then take the next one better than everyone rejected. Remarkably, this strategy succeeds in picking the true best candidate about 1/e of the time — roughly 37% — no matter how large n is, which is why it's also nicknamed the '37% rule.'
This is a genuine, well-studied mathematical result, first framed publicly in Martin Gardner's Scientific American column in 1960 and analysed extensively since. Applying it to dating is a popular, tongue-in-cheek illustration, not real relationship advice: real dating doesn't hand you a known, fixed pool size, candidates aren't strictly rankable on one dimension, and people you passed on rarely wait around to be reconsidered. Treat the result as a fun bit of math, not a life plan.
- Enter Number of people you expect to date/consider — your best guess at how many candidates you'll evaluate in total.
- Read Reject this many, then accept the next one better than all before — the count of candidates to use purely as a benchmark-setting phase.
- After that many, the rule says: pick the very next candidate who beats everyone you've seen so far, even if a 'better' one might theoretically still be ahead.
- This is a mathematical curiosity applied to dating for fun — not a substitute for actual judgment about actual people.
Worked example — sizing up 10 candidates
Enter 10 into Number of people you expect to date/consider. Dividing by e: 10 / 2.71828 = 3.6788, which rounds to 4. Reject this many, then accept the next one better than all before reads 4.
In practice, that means treating the first 4 people purely as a calibration phase — get a feel for the range of who's out there, but don't commit to any of them — then, from candidate 5 onward, accept the first one who's better than all 4 you've already seen. Following this rule across many independent trials gives roughly a 37% chance of landing on the single best candidate in the whole set of 10, the same success rate the rule delivers for any pool size.
Questions
What is the secretary problem, mathematically?
It's a decision-theory puzzle: candidates arrive one at a time in random order, you rank each one relative to everyone seen so far, and you must accept or permanently reject each before seeing the next — with the goal of maximizing the chance of ending up with the single best candidate overall. The 1/e rule (reject the first ~36.8%, then take the next record-setter) is the provably optimal strategy for this exact setup.
Why does the number e specifically show up here?
It falls out of the calculus used to find the optimal cutoff point. As the pool size grows large, the probability of success for a given rejection fraction is maximized right at 1/e ≈ 0.368, and that same fraction also turns out to be the strategy's overall success probability — a tidy coincidence that's part of why the result is so well known.
Does this actually work for real dating?
Only as a loose analogy, not literal advice. The mathematical version assumes you know the total candidate pool size in advance, can strictly rank every candidate against every other on one dimension, decide instantly and permanently on each one, and never risk a rejected candidate finding someone else in the meantime — none of which hold for real relationships. Treat this instrument as a fun illustration of optimal-stopping math, not a dating strategy.
Where does the '37% rule' name come from?
From the strategy's success probability. Reject the first n/e candidates, then take the next record-setter, and the chance of that strategy landing on the actual best candidate in the whole pool converges to 1/e — about 0.368, or 37% — regardless of how large n is. The same number, 1/e, describes both the size of the rejection phase and the odds of success.
What if I only expect to consider one or two people?
The reject count rounds very small. For 1 candidate, round(1/e) = round(0.368) = 0, meaning there's no benchmark phase at all — you simply take the only option. For 2 candidates, round(2/e) = round(0.736) = 1, so you'd pass on the first and take the second regardless of how it compares.
Is this rule used for anything besides dating?
Yes — the same 1/e logic is the standard illustration in operations research and economics courses for problems like hiring from a stream of job applicants, choosing an apartment while touring a fixed number of listings, or picking a parking spot as you approach your destination. Dating is simply the most popular pop-math framing of the underlying secretary problem.