The 37% Rule Explained: The Secretary Problem
Quick answer: The secretary problem asks how to pick the best candidate when you see candidates one at a time and must accept or reject each immediately. The optimal strategy is the 37% rule: reject the first 37% (more precisely 1/e ≈ 36.8%) of candidates, then hire the first one who beats everyone seen so far. This finds the very best candidate about 37% of the time, the best possible chance under these rules, and the chance stays near 37% as the number of candidates grows.
You are hiring for a job, searching for an apartment or, as some popular books joke, looking for a partner. Options arrive one by one. Once you pass on one, it is gone forever; once you accept, you stop. How do you maximize your chance of picking the very best? Mathematics has a surprisingly clean answer: the 37% rule.
The Problem
- There are \(n\) candidates, arriving in random order.
- You can rank each against those already seen, but you learn nothing about future ones.
- After each interview you must hire immediately or reject forever.
- Your goal is to hire the single best candidate. Second best counts as failure.
The 37% Rule
- Look: interview and reject the first \(r\) candidates, just to learn the standard.
- Leap: after that, hire the first candidate better than everyone seen so far.
For large \(n\), the best cutoff is \(r \approx \frac{n}{e} \approx 0.368\,n\), and the chance of getting the very best is then about \(\frac{1}{e} \approx 36.8\%\). For small \(n\), use the exact table below rather than rounding 37%. Here “optimal” means one specific thing: the highest probability of picking the single best candidate. It does not mean the best average quality.
Worked example with 10 candidates. Skip candidates 1, 2 and 3 (not 4: rounding 37% of 10 to 4 gives a slightly lower chance). Then take the first of candidates 4 to 10 who beats everyone seen so far. Your chance of ending up with the very best is about 39.9%. If nobody beats the first three, the strict rule ends with no hire; many people take the last candidate instead.
| Candidates \(n\) | Best number to skip | First candidate you may pick | Chance of picking the best |
|---|---|---|---|
| 3 | 1 | 2 | 50.0% |
| 5 | 2 | 3 | 43.3% |
| 10 | 3 | 4 | 39.9% |
| 20 | 7 | 8 | 38.4% |
| 100 | 37 | 38 | 37.1% |
| Very large | about 36.8% | the next one | about 36.8% |
Where 1/e Comes From
If you skip the first \(r\) candidates, the chance of success is
\[ P(r) = \frac{r}{n} \sum_{i=r+1}^{n} \frac{1}{i-1}. \]
For large \(n\), with \(x = \frac{r}{n}\), this becomes \(P \approx -x \ln x\), which is largest at \(x = \frac{1}{e}\), giving \(P = \frac{1}{e}\). The number \(e\) appears naturally; see our explainer on e.
Remarkably, the success chance does not shrink toward zero as \(n\) grows: even with a million candidates you still find the best one about 37% of the time.
The secretary problem is one model in the wider field of optimal stopping: deciding when to stop searching and act. Change the goal or the information available, and the best rule changes too.
Variations
- Happy with a top candidate, not just the best: you should start leaping earlier.
- Known scores (for example, you know scores are spread evenly between 0 and 100): you can do much better, with a threshold that drops over time.
- Candidates might say no, or you can go back to earlier ones: the best cutoff changes.
History
The problem circulated in the 1950s and was popularized by Martin Gardner in 1960. Analyses by Dennis Lindley (1961) and others established the 1/e solution, and it became a classic of optimal stopping theory, discussed in the popular book Algorithms to Live By.
Common Misconceptions
“The rule guarantees the best choice.” No. It gives the best possible chance, about 37%. About 37% of the time you end up with no one, because the best was in the rejected group.
“It is a dating law.” No. The model assumes you know how many options there will be, see them in random order, can only rank them against each other, cannot go back, and count anything but the very best as failure. Real dating breaks almost every one of these assumptions, so treat the rule as a thinking tool, not a formula.
Watch: The 37% Rule in Video
If you learn better by watching, these videos are worth your time.
Check Your Understanding
Five quick questions to see how much of this has stuck. Pick an answer and press Submit; you will see a short explanation either way.
Answer each question and press Submit to see your result right away.
Frequently Asked Questions
What is the 37% rule?
In optimal stopping, reject the first 37% of options to learn the standard, then choose the first option better than all of those. It maximizes the chance of picking the very best.
Why is the answer 37%?
For large numbers of candidates, the success probability is about −x ln x, where x is the fraction skipped. This is largest at x = 1/e ≈ 0.368.
Does the 37% rule always find the best option?
No. It finds the best option about 37% of the time, which is the highest possible chance under the problem’s rules. About 37% of the time, the best was among the rejected group and the strict rule hires no one.
Does the 37% rule work for dating?
Only as a loose idea. The math assumes a known number of options, random order, no going back and no partial credit for a great but not perfect choice. Real life rarely fits these assumptions.
What is the secretary problem?
A classic puzzle about hiring the best of n candidates seen one at a time, where each must be accepted or rejected immediately.
Keep Exploring
- What Is e in Math? Euler’s Number 2.718 Explained
- Monty Hall Problem: Why Switching Wins 2/3 of the Time
- Prisoner’s Dilemma Explained: Payoff Table & Examples
Sources and Further Reading
Related to This Article
More math articles
- SIFT Math Practice Test Questions
- 15 Free Math Websites for Students: Lessons, Practice and Tools
- The Sociological Imagination
- 5th Grade KAP Math Worksheets: FREE & Printable
- How to Match Word Problems with the One-Step Equations?
- Feet and Inches Calculator
- Vermont Algebra 1 Free Worksheets: Free Printable Algebra 1 PDF Worksheets with Worked Solutions
- The Ultimate AP Calculus BC Course
- Grade 2 English Practice for Arkansas Second Graders
- The Best Grade 8 ELA Practice Tests for Oklahoma Students


What people say about "The 37% Rule Explained: The Secretary Problem - Effortless Math"?
No one replied yet.