Love Calculator

The calculator above takes two names and returns a number between 0 and 100. It is a party trick. The number comes from the letters of the two names and from nothing else, so it carries exactly as much information about two people as their spelling does, which is none. What follows is what the number actually is, and then what the research says about attraction and about why compatibility resists this kind of calculation.

What the number actually is

The two names are stripped down to their letters, lower-cased, sorted into a fixed order and joined. That string is run through FNV-1a, a small general-purpose hash function, and the result is taken modulo 101, which yields a whole number from 0 to 100.

Three consequences follow, and they are worth knowing because they show what the score is not. Order does not matter: putting the second name first gives the same answer, which is why the Swap button never changes anything. Punctuation and capitals do not matter, so Mary-Jane, mary jane and MaryJane all score alike. And identical names do not score 100. A hash is not a similarity measure, so two people called Sam land wherever their shared spelling happens to land, which is usually nowhere near the top.

A good hash spreads its inputs evenly across its output range, which is the property that makes it useful for storing data and useless for measuring affection. Change one letter of one name and the score jumps somewhere unrelated. Score the same name against a dozen different partners and the results scatter across the whole range with no pattern, because there is no pattern to find.

Definitions of love

Love is a word doing many jobs at once. It covers the attachment between a parent and an infant, the pull between new partners, the settled companionship of decades, and the regard people hold for friends, places, work and ideas. English uses one word where Greek used several, distinguishing eros, philia, storge and agape, and that flattening is part of why any single definition feels wrong to somebody.

A common split is between impersonal and interpersonal love. Impersonal love is directed at an object, a principle or a pursuit: a craft, a cause, a piece of music, a place someone grew up in. Interpersonal love runs between people, and it is what the calculator pretends to measure.

Robert Sternberg's triangular theory, published in 1986, breaks interpersonal love into three components and remains the most useful map of the territory. Intimacy is closeness and disclosure. Passion is physical attraction and longing. Commitment is the decision to stay. Different combinations produce recognisably different relationships: intimacy alone reads as friendship, passion alone as infatuation, commitment alone as an empty marriage. Intimacy with commitment but no passion is companionate love, the shape long partnerships often settle into. All three together Sternberg called consummate love, and he was clear that holding all three at once over decades is harder than reaching it.

What attachment research adds

Cindy Hazan and Phillip Shaver argued in 1987 that adult romantic bonds run on the same system that governs infant attachment, and that people fall into recognisable patterns. Roughly, secure attachment means comfort with closeness and with being depended on. Anxious attachment means wanting closeness while fearing it will be withdrawn. Avoidant attachment means valuing independence and finding intimacy uncomfortable.

These patterns predict how people behave under stress in a relationship rather than who they are drawn to in the first place, and they are not fixed. The pairing that produces the most predictable friction is anxious with avoidant, where one partner pursues closeness precisely as the other withdraws from it. No arrangement of letters in two names has any bearing on this.

Approaching love

The findings on initial attraction are older and better established than most people expect, and they are unromantic.

Proximity does most of the early work. Leon Festinger, Stanley Schachter and Kurt Back studied friendship formation in a housing complex for married students at MIT in 1950 and found that the strongest predictor of who became close was simple physical distance, down to which stairwell a flat sat near. Being nearby, repeatedly and without effort, is how most relationships begin.

Familiarity compounds it. Robert Zajonc showed in 1968 that repeated exposure to a face, a shape or a nonsense word increases liking for it, with no need for the person to notice the exposure. Seeing somebody often makes them more appealing, which is a fair description of how workplaces and classes turn into relationships.

Similarity matters more than opposition. Donn Byrne's work from 1961 onward found that agreement on attitudes reliably increases liking, and the popular belief that opposites attract has held up poorly whenever it has been tested. People pair with others near themselves on education, background and outlook far more often than chance would give.

There is a neat historical irony here for a page about computer matching. In 1966 Elaine Walster and colleagues ran a computer dance at the University of Minnesota, pairing hundreds of first-year students by machine and measuring what predicted whether they wanted a second date. Personality tests, attitude measures and intelligence scores barely registered. Physical attractiveness, rated in a few seconds as students collected their tickets, predicted almost everything. The first serious attempt to compute romantic compatibility mostly measured looks.

What predicts a relationship lasting

Getting together and staying together are different problems with different predictors. John Gottman's observational work on married couples found that the ratio of positive to negative interactions during conflict separates stable couples from unstable ones, with stable couples running around five positive exchanges for every negative one. He also identified four behaviours that forecast breakdown: criticism aimed at the person rather than the problem, contempt, defensiveness, and stonewalling.

The broader finding across the field is that what people do together beats what they each bring. In 2020 Samantha Joel, Paul Eastwick and a large group of collaborators pooled 43 longitudinal datasets covering more than 11,000 couples and used machine learning to predict relationship satisfaction. Relationship-specific variables such as perceived partner commitment, appreciation and sexual satisfaction did real predictive work. Individual traits did much less. Most striking for a compatibility calculator: how well two partners matched on their traits added essentially nothing once each partner's own characteristics were accounted for.

That is the serious version of what this page is joking about. With eleven thousand real couples, decades of survey data and modern statistical machinery, matching people on measured traits barely predicted anything. Two names were never going to.

The one real finding about names

There is a genuine research thread on names and attraction, and its history is instructive. Brett Pelham and colleagues proposed implicit egotism: because people like themselves, they drift towards things that resemble their own names. The published claims included that people are disproportionately likely to marry others whose first or last names resemble their own, alongside the better-known claims about Dennises becoming dentists and people moving to states resembling their names.

Uri Simonsohn reexamined the evidence in 2011 and found the marriage and occupation results largely dissolved once confounds were handled, including ethnic naming patterns, reverse causation in surname data and the effect of birth-cohort name popularity. What survives is much weaker than the original claims. So the honest position on names and romance is that the one serious attempt to find an effect mostly did not replicate. A calculator that hashes two names is not a diluted version of that research. It is unrelated to it.

How to use the number

As a conversation starter, freely. As a tiebreaker, never. The score is stable, so the same two names always return the same figure and nobody can reroll a bad result, which is the only sense in which it is fair. If a low number stings, that reaction is the interesting part of the exercise, not the number that provoked it.