37% Rule Dating Calculator
Use optimal stopping math to estimate when the Look Phase should end and when the Leap Phase should begin.
Data Source
Based on optimal stopping research, the Secretary Problem, and the dating interpretation popularized in Algorithms to Live By.
Caveat
This is an educational probability model, not relationship, psychological, or life-planning advice.
37% Rule at a Glance
A strict boundary for explore vs. exploit
The 37% rule answers a simple but uncomfortable question: when do you stop collecting more dating experience and start acting on what you have learned? In the classical optimal stopping problem, the answer is to reject the first 1/e fraction of options, which is about 36.79% and usually rounded to 37%.
In dating language, that means you spend the first part of your timeline or partner pool building a benchmark, then commit to the first person who clearly beats that benchmark. The rule does not promise a perfect outcome. It tells you the strategy that maximizes your chance of choosing the single best option in a random sequence.
Quick Answer: Use the first 37% of your dating window to observe and compare without committing. After that, commit to the first person who is better than everyone you saw earlier.
The 37% rule is a practical expression of the explore vs. exploit tradeoff. Exploring means gathering more information. Exploiting means acting on the best information you already have. In dating, both errors are costly: settling too early means missing better future matches, while waiting too long means the sequence can end before you act.
The rule comes from the Secretary Problem, a probability model in which candidates appear one by one, must be accepted or rejected immediately, and cannot be recalled later. Christian and Griffiths translated that logic into human decisions by showing that the same math can be applied to time windows such as ages or to finite counts such as expected serious partners.
Relative Ranking
The model assumes you only know whether a new partner is better or worse than the people you have already dated. You do not know their global percentile in advance.
Two Phases
The rule creates a strict Look Phase and a strict Leap Phase. You do not blend them if you want the classical probability result.
Dating Translation
When you cannot estimate lifetime partner count well, you can apply the same cutoff to an age range instead of to a raw number of candidates.
The goal is narrow: maximize the chance of selecting the absolute best candidate from a random sequence. That is different from maximizing average satisfaction, minimizing regret, or accounting for emotional search costs.
The threshold comes from the constant 1/e, where e ≈ 2.71828. In the large-sample limit, skipping the first 1/e share of the sequence gives the highest possible probability of ending with the top-ranked option. That probability is also 1/e, which is why the same constant describes both the cutoff and the success rate in the classical version of the problem.
For dating, two forms are useful. The first is time-based, where you define the age window during which you are actively searching. The second is volume-based, where you estimate the number of serious relationship opportunities you expect to have.
Threshold age = Start age + (End age − Start age) × (1 / e)Reject count = round(Total partners × (1 / e))The formula is simple because the hard work is hidden in the probability proof. Your job is just to define a realistic search window, apply the cutoff, and then obey the rule consistently.
The best way to understand the model is to work through both forms. One example uses ages, because many people do not know their future partner count. The second uses a fixed number of expected serious partners, which makes the threshold easy to visualize as a sequence.
Define the search window
Suppose you started seriously dating at 18 and want to be settled by 35. That creates a total search span of 17 years.
Apply the 1/e constant
Multiply the span by 0.36788. The explore portion is 17 × 0.36788 = 6.2540 years.
Translate the threshold back into age
Add the explore span to the starting age: 18 + 6.2540 = 24.2540. Rounded to two decimals, the threshold is 24.25 years.
Interpret the partner-count version
If you instead expect 10 serious partners, the rule says round(10 × 0.36788) = 4. Partners 1 through 4 are purely informational, and the Leap Phase begins with partner 5.
Age Interpretation
A result of 24.25 means the switch happens at roughly 24 years and 3 months, not at the end of age 24.
Partner Interpretation
A reject count of 4 does not mean “choose partner 4.” It means partner 4 finishes the benchmark set, and partner 5 is the first candidate you may accept.
The result is only useful if you interpret it as a behavioral boundary, not as a vague suggestion. During the Look Phase, your task is to observe patterns, compare experiences, and build a benchmark. During the Leap Phase, your task changes completely: accept the first person who beats that benchmark rather than continuing to sample.
For finite partner counts, the exact success probability is often slightly higher than the asymptotic 36.79%. Small samples produce discrete cutoffs, which is why a pool of 10 partners yields a 40/60 visual split even though the continuous rule is “37/63” in the limit.
| Total Expected Partners | Look Phase | Leap Phase Begins | Best-Match Probability |
|---|---|---|---|
| 5 | Reject 2 | Partner 3 | 43.3% |
| 10 | Reject 4 | Partner 5 | 39.8% |
| 20 | Reject 7 | Partner 8 | 38.4% |
In the classical rule, if you reach the end of the sequence without seeing anyone better than your best Look-Phase candidate, you accept the final option. That unpleasant boundary condition is part of why the model feels strict but mathematically coherent.
The standard dating version assumes a no-recall environment with only relative ranking. Real life often violates those assumptions. Once that happens, the exact 37% cutoff is no longer sacred. The direction of the correction depends on what kind of extra information or flexibility you have.
These variants are useful for interpretation, but this calculator intentionally implements the classical 37% rule rather than a custom search-cost or recall model.
Optimal stopping is elegant because it turns uncertainty into a rule. It is limited because human relationships are not random, one-sided, cost-free sequences. The model is still useful as a thinking tool, but it becomes misleading if you treat it as a literal script for major life decisions.
No-Recall Assumption
The classical model assumes that once you pass on someone, they are gone forever. Human dating often includes delayed reconnection, which changes the threshold.
Mutual Consent
The secretary problem is one-sided. Dating is not. The person you choose can reject you, which means the real-world optimization problem is harder than the textbook version.
Apps Change the Information Set
Dating apps provide filters, bios, photos, and algorithmic sorting before you even meet. That violates the pure no-information setup used by the classical proof.
Best Option vs. Good Option
The rule maximizes the chance of selecting the single best option, not the chance of choosing a merely good, stable, or emotionally compatible partner.
Treat the result as a structured way to think about decision timing, not as an instruction that should override values, safety, readiness, or mutual compatibility.