Why the 37% Rule Is Mathematically Wrong for Choosing a Partner

The 37% rule is a beautiful piece of optimal-stopping math applied to the wrong problem. It gives smart men permission to feel rigorous about a decision that requires a different mode of analysis entirely — and the rigor it provides is the cope that keeps them single into their forties. This piece is the structural read on what the math actually says, where it breaks the moment it touches a real relationship, and what algorithm to run instead.

I'm going to do something slightly unusual here: take the math seriously before knocking it down. The 37% rule isn't dumb. It's correct for its problem. It just doesn't solve yours. Knowing why is the first step in not letting it run your selection process by default.

What the math actually says

The 37% rule is the popular name for the optimal solution to the secretary problem — a classic puzzle in optimal stopping theory. The setup, formally:

You have N candidates to interview in random order. After each interview you must decide immediately to hire or reject — no recall, no going back. You can rank candidates only ordinally (better than, worse than) relative to those you've already seen. Your goal: maximize the probability of hiring the single best candidate.

The optimal strategy: reject the first N/e candidates (~37% of N, since 1/e ≈ 0.3679), remember the best of that group, then hire the next candidate who beats that best. The probability of selecting the actual best candidate using this strategy approaches 1/e ≈ 37% as N grows.

That's it. That's the rule. It's elegant. It's provably optimal under those constraints. It is also doing very specific work that bears almost no resemblance to selecting a wife: selecting the single best candidate, under a known N, with no recall, on full ordinal information.

The popular-culture version often gets even this wrong. People say things like "date around for the first 37% of your dating life, then settle down with the next person who's better than anyone you've dated before." That formulation introduces additional errors on top of the original misapplication. We'll get to those.

Why smart men reach for it

The 37% rule is appealing to high-output men for the same reason it's misapplied: it feels rigorous, it produces a clean number, and it lets them apply analytical thinking to a problem that resists it.

A man who runs companies, builds models, optimizes portfolios, reads scans, runs cases, or closes deals wants to bring the same analytical toolkit to his romantic life. The toolkit produces wins everywhere else. Why not here? The 37% rule offers a bridge: an actual mathematical theorem about decision-making under uncertainty, with a famous solution, an easy-to-remember number, and a respectable academic pedigree.

It scratches a real itch. The itch is the discomfort of making a major decision without a quantitative framework. The 37% rule provides the framework. The framework is wrong. But the relief it provides is real, which is why so many smart men adopt it without examining whether it actually applies.

The deeper appeal is permission. The rule tells you: you should spend the first portion of your dating life exploring, not committing. For a high-output man in his late twenties or early thirties, with career consolidating, income rising, and options expanding, that's a structurally satisfying message. The math is giving him permission to do what he wanted to do anyway. That should be the first warning sign about its application.

Where the math breaks the moment it touches dating

Six specific assumptions of the secretary problem fail in actual relationship selection. Each one is load-bearing for the optimality proof, and each one is structurally wrong in the dating context.

1. Known N. The secretary problem assumes you know the total number of candidates. In dating, you don't. You don't know how many serious relationship candidates you'll meet in your remaining selection window. The whole 37% calculation depends on a denominator you cannot compute. People estimate — I'll seriously consider maybe twenty women in my dating life — but the estimate is fictional. You're stopping at the 37% point of an imagined N, which means you're stopping at no defined point at all.

2. Strict sequence and no overlap. The secretary problem assumes candidates arrive one at a time, each decision made before the next arrives. In dating, candidates overlap. You're sometimes dating two women simultaneously. The decision about one is not made in isolation from another. The "sequential, one-at-a-time" structure that makes the math tractable doesn't describe real selection.

3. No recall. The classic secretary problem forbids returning to rejected candidates. In dating, recall is sometimes possible. The optimal strategy under partial recall is different from the optimal strategy without recall, and "37%" is no longer the right number. The popular rule papers over this.

4. Perfect ordinal ranking on first observation. The secretary problem assumes you can rank each candidate accurately the moment you see her, relative to those you've already seen. In dating, your rankings update for months or years. You think she's a 7 in week three and a 9 by month six — or a 9 in week three and a 4 by month six. The information that produces accurate ranking arrives well after the decision point the rule wants you to commit at.

5. Static utility. The secretary problem assumes each candidate has a fixed value. In dating, the "value" of a partner is a function of which version of you shows up to be in the relationship. Layer 2 covers this in detail. The "best" candidate isn't a property of her. It's a joint property of her configuration and your Layer 2. The optimization target isn't well-defined.

6. Single decision-maker. The secretary problem assumes you're selecting from a passive pool of candidates who don't get a vote. In dating, she's also selecting. The pool you face has already been filtered by who's interested in you, and any candidate you "select" can decline. This is sometimes called the mutual selection problem, and it has a different optimal solution that doesn't reduce to a clean percentage.

Six failed assumptions. The optimality proof for the secretary problem requires all six. Strip even one, and 37% stops being the right answer. Strip all six, and the entire mathematical apparatus is operating on a problem that doesn't exist in your life.

The math isn't wrong. It's applied to a problem that doesn't exist in your life.

What the rule actually produces in practice

When high-output men adopt the 37% rule as a working heuristic — formally or informally — three failure modes show up.

Permission to over-explore. "I'm still in my 37%" becomes a license to defer commitment past the point where it makes sense. Since N is unknown, 37% of an imagined N can be inflated indefinitely. The rule encourages a specific kind of decision avoidance dressed up as analytical rigor. A man optimizing the secretary problem's expected value is a man who can rationalize dating until forty-five without anxiety.

Ranking paralysis. The rule asks you to identify the "best of the first 37%" and then beat it. This requires you to maintain a running ranking of every serious connection you've had — a ranking that, by the time you're in your mid-thirties, is doing more harm than good. You're not deciding on whether the woman in front of you passes a structural test. You're deciding on whether she beats a four-years-ago benchmark that has no business being in the calculation.

Optimisation against the wrong target. The rule optimizes probability of selecting the best candidate. In long-term partner selection, you don't need the best candidate. You need a candidate who passes structural verification. Those are different optimization targets. The first is unanswerable in practice. The second is testable in months. The rule pushes you to spend years answering an unanswerable question instead of running the answerable one.

The deeper category error

The 37% rule is a ranking algorithm. The Selection Standard is a filtering algorithm. They are solving structurally different problems.

A ranking algorithm answers: given a population of candidates, which one is best? It requires a comparison function across candidates and a procedure for converging on the top of the distribution. The 37% rule is a beautiful ranking algorithm.

A filtering algorithm answers: does this specific candidate meet the structural criteria? It doesn't compare her to other candidates. It compares her to a written specification. She passes or she fails. If she passes, the algorithm has produced its output. There's no "but is she the best?" because the algorithm isn't trying to answer that question.

For long-term partner selection, filtering is the correct algorithm. Three reasons.

Filtering scales with ambiguous N. You don't need to know how many candidates exist. You only need to know whether the current one passes. Filtering is robust to the central failure of the secretary problem's application.

Filtering produces stable decisions. A ranking algorithm rerates everyone every time a new candidate arrives. A filter doesn't. Once a candidate passes your filter, she passes. New candidates entering the pool don't downgrade her by comparison.

Filtering optimizes for fit, not maximization. A marriage that lasts requires structural fit. It doesn't require that your wife be the highest-ranked woman you'll ever meet. Filtering acknowledges this. Ranking ignores it.

The rest of this site is the implementation of the filter algorithm. Hard Filters define the structural criteria. The Pattern Constant diagnoses which filters are load-bearing for you specifically. The Either/Or System and SAI+ verify the filters in practice. Midchalance is the internal posture that holds the filter steady when chemistry would otherwise override it.

None of this requires you to know N. None of it requires you to maintain a ranking. None of it produces the "I'll know her when I see her" comparison fog that the 37% rule, applied to dating, generates as a byproduct.

What to run instead

The algorithm in one paragraph:

Write your Hard Filters before the next first date. Five to seven items, each binary, each structural. Apply them upstream of attraction — preferably by date two, certainly by date five. If she passes, run deeper structural verification using the Either/Or System and SAI+, looking at conflict regulation, structural reciprocity, engine-fit, pre-existing self-knowledge. If she passes those, your selection process has produced its output. Commit. If she fails any of them, the algorithm has also produced its output. Move on without ranking her against the alternatives, because ranking isn't what you're doing.

That's the algorithm. No clean percentage. No poster-worthy theorem. The correct algorithm for the problem you actually have.

The Selection Standard book is the long-form version of this algorithm. The 30 chapters cover the filter criteria, the verification methods, the failure modes, the internal postures that hold the system together when behavior pushes you off it. There is no math like the 37% rule's in the book, because the problem doesn't have a clean closed-form solution. What it has is a working procedure, and a working procedure is what you actually need.

The honest sentence

The 37% rule is a beautiful piece of math applied to the wrong problem. The right problem is not optimal stopping. The right problem is structural verification, run against a written filter, with a binary decision at the end. The math you're looking for isn't a theorem. It's a checklist — and the discipline to refuse to override it with vibes.

The men who run their dating lives on the 37% rule produce predictable outcomes: long dating windows, late commitments, and a sense that the woman they eventually marry was "the best one out of the ones who were around at the right time." That last phrase is not the language of someone who has selected. It's the language of someone who has sorted. Sorting is the wrong algorithm.

Stop sorting. Start filtering. The first move is the Hard Filter list, written before the next first date.

---

Related:

• How to Choose the Right Woman: The Selection Problem Smart Men Keep Getting Wrong — the pillar piece, where filtering is named as the structural alternative.
• Hard Filters vs Preferences — the filter algorithm in full.
• The Pattern Constant — the diagnostic that tells you which filters to weight heaviest.
• The 95% Rule: Why Most High-Output Men Misread Tony Robbins on Dating — Layer 2, which the secretary problem's static-utility assumption ignores.

The system end-to-end: The Selection Standard — a 30-chapter decision framework for choosing the right woman without burning years on the wrong ones. Hard Filters (Chapter 12), Either/Or System (Chapter 15), SAI+ (Chapter 17), Midchalance (Chapter 6). Read more →