Rules & Frameworks

“She thinks she's a 9” is one of the most-repeated lines in the manosphere, and it is pointed at women. The measured pattern does not cooperate with that aim. Miscalibration concentrates at the bottom of the distribution — in both sexes — and the top under-rates itself slightly. The claim is directionally right that a systematic gap exists and wrong about who carries it and which way it runs.

Joel, Eastwick, and Finkel gave machine-learning models more than 100 pre-date traits and preferences in two speed-dating samples (n=163 and n=187). Those models predicted general chooser and partner tendencies but explained effectively none of the held-out unique desire for this particular person (less than 0.1%). The after-date impressions — including perceived chemistry — were collected alongside desire, so they show what becomes observable after interaction rather than a prospective causal effect. A later synthesis across three longitudinal speed-dating studies (n=559; more than 6,600 initial dates) found that unique liking and consensual desirability at the first meeting predicted later interest, messaging, and dates. Tier 2

The studies establish a prediction boundary, not a romantic mystery. LE’s inference is the workflow above: let facts remove known bad fits, then let observed dyadic evidence decide close calls. This is also the boundary on the site’s own calculators: they estimate declared fit and constraints conditional on their inputs; they do not predict chemistry.

Across five independent samples of single people, Hadden, Agnew, and Tan found that commitment readiness tracked interest and active pursuit. In pooled longitudinal data, 116 people entered relationships while 529 remained single; higher baseline readiness predicted entry (OR 1.46 per one-SD increase in within-sample standardized readiness, 95% CI 1.18–1.81). Among the smaller entrant group, earlier readiness also predicted later commitment, satisfaction, and investment. But the individual longitudinal samples were not uniformly significant, the measures were self-reports, and the samples leaned U.S., young, and undergraduate. Readiness is a real predictive state, not a causal switch or a universal personality trait. Tier 2

The 7–7 isn’t an absolute — it is the floor for the long-term tier specifically. The same two axes gate at different heights depending on what is being selected for: short-term selects hard on the physical channel and forgives much of the rest; long-term raises the bar on both and adds fit. This is why the rule feels true in one context and wrong in another — “I’ve seen low-attraction relationships work” and “people hook up with personalities they’d never date” are both true, because they clear different tiers’ floors. The error that manufactures most of the guilt is applying one tier’s floor to another tier’s decision. So the thing to hold isn’t the number — it is the principle the number is one instance of: floors are tier-relative, and the 7–7 is the long-term case.

A floor is not a preference you can talk yourself past. Take the real case: someone walks in already below their physical floor and deliberately chooses to override it — gives the personality its chance, does the virtuous thing — and the attraction still never arrives. That is the gate firing as designed: it operates underneath the decision to try, not as a verdict reached by weighing. The guilt — “I felt shallow for not giving him a chance” — assumes attraction was a choice made wrongly. It was never a choice. Trying was never going to work, because the floor isn’t a moral test you pass with effort; it is a threshold that clears or doesn’t, regardless of intent. Naming the tier is what removes the shame: not shallow — the long-term gate didn’t fire.

“Is an 8 in looks worth a 6 in personality?” is half a trap. Trades only work above the floors. Below a floor there is nothing to trade, because the gate has already shut — a high score on one axis can’t reach across and lift the other over its own floor. So an 8 personality with a 6 in looks still fails, and a single charming trait — a 9 in humor — can’t rescue a composite everything else has dragged to a 6. The real question is never “does it net out,” it is “did both gates clear.”

None of this is perfectionism — it is refusing to build on a missing pillar. Sustained desire and a genuinely good personality both have to clear the bar independently.

The 7–7 Rule isn’t a cosmic constant — it’s a discipline you impose on your own ratings. Looks draw real consensus — people broadly agree who reads as a 7 versus a 4 — but where you set the passing bar is your call, and the personality composite is softer and more personal still. That doesn’t weaken the rule, it locates it. Its whole value is that you hold your own line and refuse to talk yourself beneath it — sustained desire and a genuinely good personality, both clearing the bar you actually set.

Not every trait gates — only the important ones, and ‘important’ has a precise meaning: a channel factor, one the whole experience routes through. A 4 on a channel factor doesn’t sit quietly in the average — it contaminates everything next to it. Hygiene at a 4 poisons proximity, so the 9 in humor never lands; honesty at a 4 poisons trust, so every other virtue is read through doubt. That is why it gates: you can’t reach the rest through the barrier it puts up.

This is why the gated set is small: a trait is top-tier because it’s a channel — ‘important’ and ‘channel-type’ are nearly the same set. That is the answer to “a 4 in what?”: not any nitpick — the load-bearing few.

Push the tier-relativity from the 7–7 down to the component level and the gate splits by tier — each tier gates hardest on the channel it actually runs on.

The Compatibility Calculator models the long-term case but applies the blunt form of this gate: it breaches on any factor below 5, across every tier, without yet sorting channel components from additive goods. So it runs deliberately conservative — it can flag an additive good (common interests, say) as if it were a load-bearing failure. The refinement above is the doctrine; the calculator is its strict approximation.

What about the other channel — a personality flaw short-term, a looks flaw long-term? One line settles it: the off-channel gate fires only when the flaw still damages this tier’s actual medium.

The split that matters most sits one level up. For women, safety isn’t another personality component — it’s a meta-gate above the whole structure, evaluated first, because the cost asymmetry is categorical. Threat signals (boundary-testing, volatility, disregard) don’t merely lose short-term leeway; they’re assessed before looks or anything else is scored, and a 9 on every other axis doesn’t clear a safety fail. For men the physical channel gates harder and earlier, and the safety meta-gate is largely absent or far lower-stakes, so the leeway structure above applies more straightforwardly.

This re-explains the deliberate-attempt case from the 7–7: she could afford to gamble below her looks floor precisely because the safety gate was already cleared — women’s long-term looks floor can flex partly because it isn’t the gate carrying the existential downside. So the gender difference isn’t ‘same factors, different weights’ — it is different gate architecture.

This floor runs on your own eyes. Agreement on looks is high but not total — a real slice of attraction is private taste, so a 4 to you can be a 6 to someone else — and “below 5” means below the line you draw, not a grade stamped on a person. So the rule is personal twice over: you set the scale, and the scale tracks where you sit. If you’re a sub-5 yourself, parity is the realistic pairing — another sub-5 isn’t a breach, it’s your league. The auto-disqualification is the standard as applied looking down from at or above the floor; it never obliges you to reject your own level. See SMV Matching for why people pair near their own band.

Above the floors, the two numbers add. A 6–5 (looks-6, personality-5) and a 5–6 both total 11 — and both clear every floor — so they sit at the same parity, a likely pairing. The 7–7’s no-trade rule governs qualification (clearing the ≥7 gate), not matching, so nothing forbids them. They aren’t identical — the higher-personality 5–6 carries a little more perceived value to the 6–5 — but that’s a shading within the pair, not a barrier to it.

Parity sets the bracket; treatment moves you within it. Your behavioral value-add shifts your effective total — enough to firm up a near-parity pairing, or to reach a half-step above your looks-and-personality line. So parity isn’t a static lattice of coordinate-twins; it has give.

Two 6–2s (looks-6, personality-2) pair at parity — 8 each, no value gap between them. They break the Sub-5 floor, yet they’re each other’s realistic match. That isn’t a loophole — it’s the rule resolving a conflict, because Sub-5 and Parity bind two different demographics.

How wide is “your tier”? The Attention Market’s reachable bands draw the lines. For a 5, the realistic commitment band is about 4–6 (±1), with an outer reach to ~7 only when status and fit carry it — past that, the pairing is out of bounds and won’t hold.

If pairing is parity, why does everyone feel they reached? Because the reach is real — it just lives in the Option Pool, not the marriage. Attention and short-term access genuinely run above your line; commitment regresses to parity. That gap — between who you can briefly attract and who you actually keep — is exactly the space the Spiderman Effect fills.

The lift isn’t a switch; it’s a probability that decays with the gap. And you don’t have to take that on faith — it is the upgrade tail of the Looks Rating drift model, the same frequencies read from the other side.

Honest caveat: the Looks Rating model measures one rating drifting over time (and only splits “+1” from “>1”), not pair-bonded elevation — so the +2 and +4 figures are an extrapolation of that tail, the shape borrowed, not a direct measurement. But the message holds: +1 is a plausible minority, +2 is rare, beyond that is delusion.

The bounded curve is what keeps this from being a license to ignore the gates. Inside the parity band, mutual elevation is a stable equilibrium — two near-equals each genuinely experiencing the other as above baseline. Stretch it across three tiers and the same belief is just the soft sham telling itself a story.

It also answers what the Status Trade only half-settles. Women’s desire tilts up; the Option Pool gates the up-conversion, so pairing lands at parity anyway. The Spiderman Effect is what makes that bearable — her parity partner reads as the up-match she wanted. Hypergamy (what she wants) + parity (what she gets) + Spiderman (what reconciles them) is the closed loop — and it’s why a tilt that never converts still produces stable, satisfied pairs.

A coordinated strike is a collective-action problem with the worst possible incentive structure. The payoff to defecting — being the one man who keeps approaching while the others abstain — doesn’t merely stay positive; it rises the more everyone else complies. A near-empty field is a target-rich one. The equilibrium is unstable by construction: the harder the strike holds, the more it pays to break it, and someone always does.

This is a structural argument, not a measured one — but the structure is the point: the strike fails as a lever before any data is collected. Lens

Nobody had to organize anything. The marginal cost of initiating rose — rejection risk, ambiguous norms, and apps that funnel the median woman’s attention toward the top decile of men, so the average approach pays worse than it used to — and men individually, silently, dropped out. Same outcome as a strike, with zero coordination required. It is an attrition, not a refusal.

The hypothetical’s load-bearing claim is that some fraction of women, feeling the absence, would start initiating — signalling harder, approaching, raising their own effort. This is the weakest leg, and the data cuts against it. Women’s dominant response to a thin market is to leave it, not to compete in it.

So the optimistic “40% start changing” is the part to discount. The realistic split isn’t cope-vs-initiate; it’s exit-vs-wait. Hunch

We don’t have to imagine the strike; two real movements have run a version of it, one per sex. They are the closest thing to a controlled test, and the verdict is clean: the demographic facts are Tier 1, but the claim that either movement caused the decline is Tier 3 and largely unmeasured.

The lesson for the strike thesis: when withdrawal is actually attempted, the aggregate numbers don’t bend to the strikers’ will. That is the best evidence we have that the lever is weak. Lens

The canonical demonstration is Iyengar & Lepper’s jam table: the 24-flavor display drew more browsers, but the 6-flavor display converted ten times the buyers (30% vs 3%) — and restricted-set choosers reported more satisfaction with what they picked. Schwartz built The Paradox of Choice on this family of results, adding the useful split between maximizers (must find the best option, and abundance punishes them) and satisficers (need good-enough, and abundance mostly doesn’t). Tier 2

The argument’s spine is contraceptive. Hormonal birth control — Enovid, FDA-approved 1960 — gave women reliable unilateral control over conception, with no male equivalent developed in the sixty-five years since. Before it, both parties negotiated the terms of sex at the point of sex; after it, one party could privately decide the outcome. Tier 1 as history: the date and the asymmetry are real and undisputed.

Two anecdotes: an impression of sitcom casting, and a single radio call-in about a wife who deceived her husband about contraception. Neither is data. That does not make the structural claim false — incentive asymmetries can be real with no study behind them — but it means holding this exactly where the site holds every unmeasured mechanism: useful as a lens, not laundered into a finding.

Burdett and Coles built a marriage-market model where people differ on a single shared ranking of desirability, everyone prefers a better partner to a worse one, both sides must accept, and — critically — nobody has any preference for similarity at all. Add one ingredient: searching takes time, and time is costly. The equilibrium that falls out is that the market partitions itself into classes, each accepting only its own contiguous band, because holding out for better costs more than accepting now. Tier 2

The discourse moved from “the apps are inevitable” to “the apps are dying” in about three years. The measured version is smaller than either slogan.

Spence's job-market signaling model gives the exact condition, and it is narrower than the slogan: a signal separates strong senders from weak ones when its cost differs across them — cheaper for the strong to send than for the weak. What does the work is the difference in cost between types, not the absolute expense. Zahavi's handicap principle is the same logic in biology: the display is credible because it is differentially costly. Tier 1 as economic theory — Spence shared the 2001 Nobel for this class of result.

Toma, Hancock and Ellison did the unglamorous thing: they brought 80 online daters into a lab and established ground truth for height, weight and age against what those daters' profiles claimed.

Courtship used to run through people who already knew you. That network was an expensive signal in itself: a friend who vouched for you staked their own standing, and a reputation built over years could not be assembled the night before. Moving courtship online did not remove the need to verify anyone — it transferred the cost from a community to an individual, who now has twenty minutes and a coffee to do what a village used to do continuously.

Sprecher and Felmlee followed couples across three waves and found the plain result: relationships with more support from the surrounding network reported more love, satisfaction and commitment, and were less likely to have broken up later. Approval from the woman's network was the stronger predictor of the two. Tier 2

The Romeo and Juliet effect is one of the most-repeated findings in popular relationship psychology: parental interference increases love. It entered the literature in 1972 and entered common sense shortly after. When it was properly re-examined — 396 participants tracked over three to four months on the original measures — it did not appear. Higher interference and lower approval predicted poorer relationship quality on every outcome measured. Tier 2

This site already carries both halves of the change and has never joined them: the collapse of meeting through friends and family, and the decline of the third places where that used to happen. The usual reading is nostalgic — something warm went away. The structural reading is colder and more useful.

Across four studies, Gable and colleagues linked active-constructive responses to positive-event sharing with greater daily positive affect, satisfaction, intimacy, trust, and commitment. The designs included student diaries, 59 heterosexual dating couples, and 89 married couples followed in 14-day diaries. A separate observational study of 79 dating couples found that responses during videotaped positive-event discussions related more strongly to well-being and two-month stability than responses to negative-event discussions. The convergence is useful; the evidence is still largely correlational, self-reported, and concentrated in one research program. Tier 2

Sex-only, the inheritance marriage, the green-card arrangement, status cover. The defining variable is intent — and intent isn’t on the looks or personality axes, so a hard sham is unplottable from the numbers. The scores can read as a flawless 8/8 and be entirely beside the point, because the relationship was never about the attraction they measure.

So this layer needs a taxonomy of motives, not a graph — you can’t chart intent on attraction axes any more than you’d chart a tax bracket on height.

Both partners can clear their floors and the pairing still be a sham — the negative-space mirror of the Spiderman Effect, where mutual belief can’t hold the gap. This layer lives in score space, and bundles two mechanically distinct cases that behave alike — they don’t last, or they last begrudgingly:

The soft layer is the plottable one — but its two cases live in different spaces. Rule violation shows on an individual gate plot: looks-composite on one axis, personality-composite on the other, where the 7–7 draws a right-angle viable quadrant and Sub-5 punches holes inside it — a violation is a pairing that persists with a partner sitting outside those boundaries. Out-of-league mismatch shows on a couple plot: your looks against your partner’s, where the Matching Curve — not the parity diagonal — marks the expected pairing, and a mismatch is a lasting pairing sitting far off that curve. One lens, two coordinate systems.

Aguiar, Bils, Charles and Hurst asked why young men's working hours fell so much faster than everyone else's. Between 2004 and 2015, young men's video-game time rose by roughly 99 hours a year — about 50%. That descriptive time-use trend is solid. Tier 1

Substitution and complementarity are observationally identical in cross-section. The man who games because dating went badly and the man who stopped dating because gaming is better generate the same time-use row, the same survey answer, and the same anecdote. Nobody has separated them.

The number is manufactured by dividing this year’s divorces by this year’s marriages. The numerator is drawn from every marriage currently in existence — couples who married across sixty years of cohorts. The denominator is the couples who married this year. By construction the two groups share no members. Nobody counted in the numerator is in the denominator. It is not a probability of anything.

The tempting version of this doctrine is “snapshots make temporary states look permanent,” and it is wrong often enough to be dangerous. Kennedy and Ruggles found the opposite failure in the same subject: the unstandardised divorce rate looked flat from 1980 onward, while the age-standardised rate rose 40%. Every age-specific rate was climbing; the aggregate hid it, because the married population had aged out of the high-divorce years. Tier 1

Never-married-at-40 is a stock, and reading it as a lifetime outcome assumes the first-marriage rate after 40 is zero. It is not zero, and it is the fastest-growing part of the distribution.

No primary source states this. Not a study, not a platform, not a dataset. Tracing it is worth doing in full, because it is the most-cited statistic in the manosphere and it is a composite of two unrelated claims, neither of which says what the slogan says.

Akerlof’s used-car market is the canonical form: when buyers cannot tell good from bad, the bad drive out the good, because both sell at the same price. The engine is a one-sided information asymmetry — in his words, only the seller knows the difference. Tier 1 as economic theory; Akerlof shared the 2001 Nobel for it.

The unpartnered share of Americans aged 25–54 went from 29% in 1990 to 38% in 2019, and all of the growth was in the never-married. Against partnered people of the same ages, the differences are real and they are not symmetric between the sexes — which is the fact that decides how this entry has to be read. Tier 1 (Census/ACS microdata via IPUMS)

The inference the discourse draws is that exits happen in quality order, so what remains is the bottom of a queue. Each step of that fails against a different body of evidence, and the first failure is already doctrine on this site.

Federal survey data on the probability of a first marriage by a given age, broken out by education, produces a gradient that is the reverse of the folk model — and then reverses again.

First marriages per 1,000 never-married people, by age:

The measurable transformation of the pool is not that the top is removed. It is that the pool stops being made of people who were never married and starts being made of people who were. Previously married — divorced or widowed — as a share of all unmarried adults:

Effective supply = local pool × age fit × availability × active search × accepted economic band. Every term is a filter. Change one and the apparent surplus can shrink, disappear, or reverse.

Researchers cannot randomise the sex ratio, so the credible work leans on catastrophes and migration flows. Four independent designs, none correlational:

The argument is always conducted about places. Across US metros, the ratio for 25–34s is nearly flat: a median around 102 men per 100 women, a standard deviation of about 5, and a full range of roughly 91 to 116. Apply the instrumented coefficient above to a move from the very worst US metro to the very best and it buys about 3.7 points of marriage probability — real, measurable, and nowhere near a change of regime. Now condition on a credential instead. Men per 100 women, ages 25–34:

Unmarried men per 100 unmarried women, by age. The left column counts only the never married; the right adds the divorced and widowed. Same country, same census, same moment.

Behavioural ecology has had the right concept since 1977: the operational sex ratio counts individuals actually competing to mate, not individuals who exist. Human demography never built the equivalent — its refined measures adjust for who is suitable (age, race, education), never for who is searching.

A single adjusted number would be false precision, because it inherits every weakness of the input. The defensible object is the threshold, which needs only the census counts:

An IFS/YouGov survey of roughly 3,000 U.S. adults aged 18–29 (April 2025) asked how important shared political views are in a life partner. The answer is not one number; it is four, and the spread between them is the entry.

The honest consequence of this entry is that it eats its own tail, and the site should say so rather than stop at the flattering level of resolution.

It is the obvious action, it is what the popular treatments of this subject recommend, and it has been tested directly. It does not survive the test.

Each of these tests the “everyone found out” story directly. All three fail it.

Before attributing a fading effect to saturation, it has to beat two nulls that involve no diffusion whatsoever.

The everyday form of this doctrine is “everyone recognises that move now, so it stopped working.” Recognition has been measured, and it does not do that.

“Nothing has been tested” would be false, and the honest exceptions are the most useful evidence in the field.

This entry started from “the failures are invisible.” The evidence supports something more specific and more damning: the failures are frequently right there in the source, and they do not survive the journey to the advice.

Advice content is often not missing citations. It has things that look exactly like citations and are not.

An entry that only sneers would be a worse entry, and would also be wrong.

The natural version of this entry says an engagement-optimising feed selects against accuracy. We wrote that version first, and then went looking for the test.

Advisors were manipulated on two axes, confidence and accuracy, and audiences rated their credibility and took their advice. The result depends entirely on whether the audience gets to see who was right.

The satisfying conclusion is “if it went viral, it is probably false.” That does not survive contact with the numbers, and this site would rather publish the boring bound than the quotable slogan.

The same turn runs through the wider clinical vocabulary, and one test separates its honest use from its weaponised one.

Becker’s 1973 model is the source of every “marriage is about specialisation” argument you have ever read. Its gains come from comparative advantage, which yields one distinctive, falsifiable prediction: mates should sort negatively on wage rates. Positive sorting on schooling, height, intelligence and the rest is predicted by half a dozen models with no specialisation in them at all. The wage prediction is the one that is his.

The standard successor story: as women’s wages converged, gains from production gave way to gains from consumption — shared tastes, shared leisure, companionship. We went looking for the measurement behind it.

There is exactly one literature that identifies the two questions separately: a consumption technology (how much a household produces from given resources — the size) and a sharing rule (who ends up with what — the split). It is worth knowing how differently the two halves have fared.

Everything above comes from places where leaving is genuinely available and the leaving is enforced. Where it is not, the same shock has been observed to run the other way. This is not a caveat; it is the boundary of the mechanism.

This entry was drafted around a well-known finding that appears to invert it: wives who out-earn their husbands do more housework, not less, which is the exact opposite of what bargaining predicts. It is a genuinely famous result and we intended to give it the last word. Then we read the papers that came after it.

Everything above is about households. Every Tier 1 result in this entry needs one of two things: a legal regime governing exit, or an observable pooled budget. Dating has neither, and there is no equivalent literature. This is the honest position and it is worth more than a confident extrapolation.

American states adopted unilateral divorce at different times over three decades, which is as close to a controlled experiment on the price of exit as this subject will ever get. What it shows is not what either side of the culture-war version expects.

Three American states built exactly the device this entry describes and offered it to everyone. Covenant marriage is a real, legally binding, harder-to-exit marriage contract, available at the same price as the ordinary one. It is the cleanest test of whether people want commitment technology.

An entry about the loss of binding commitment can be read as nostalgia for it. That reading has to be closed off with evidence rather than with a disclaimer, because the evidence is specific and it is serious.