15 July 20269 min readRelationships, Decision-making

The Wrong Insurance Policy

The dating advice to pick the one who'll be there on your worst day sounds wise. It optimizes for a scenario that happens a handful of times in a life while ignoring the trait that predicts every ordinary Tuesday.

There's a piece of dating advice that circulates on TikTok in various forms, always delivered with the moral clarity of something everyone already agrees with: don't pick the tall one who gives you butterflies. Pick the one who'll be with you on the worst day of your life.

It sounds right. It sounds like the kind of thing a person who has been through something would say. And every time I see it, I think the priority ordering is backwards.

Looks don't matter more than character. That's the strawman version of what I'm about to argue, and it's wrong. What I'm arguing is that the advice misidentifies which signals are reliable, overweights a scenario that almost never occurs, and tells you to prioritize the one trait your partner can develop while ignoring the one thing they can't.

Most of life is Tuesday

The advice is built on a framing that privileges the exceptional over the ordinary. "The worst day of your life" is, by definition, a tail event. It's rare. For most people, it happens a handful of times across a full lifespan: a death, a diagnosis, a financial collapse. The rest of life is normal. Boring, even. Tuesday mornings, commutes, cooking, grocery runs, sitting on the couch deciding what to watch. If your relationship works on those days, you will be happy in that relationship most of the time. If it doesn't, no amount of crisis solidarity will fix it.

This isn't speculation. The research on relationship satisfaction overwhelmingly tracks daily, dispositional factors, not crisis response. A meta-analysis of the Big Five personality traits and relationship satisfaction, Malouff, Thorsteinsson, Schutte and colleagues, aggregating 40 studies and over 7,600 participants, found that the strongest predictors of marital satisfaction were emotional stability (r = .29), agreeableness (r = .29), and conscientiousness (r = .25). These are dispositional traits that show up on Tuesdays: how pleasant you are to be around on an average day, how reliably you follow through, how little unnecessary friction you generate. None of them measure crisis behavior. "Will be with you on the worst day of your life" isn't a Big Five factor, because it isn't what makes relationships work across the years that contain no catastrophe.

O'Meara and South (2019, Journal of Personality) followed married adults longitudinally and found that changes in neuroticism, the inverse of emotional stability, tracked in parallel with changes in relationship satisfaction over time. Relationship happiness moved with daily emotional regulation, not with some stored-up reservoir of crisis capacity that gets deployed when things go wrong.

The advice tells you to optimize for the catastrophe payout when you're buying a daily-use product.

The signal problem

Here's a harder problem with the advice: the trait it tells you to select for is nearly impossible to assess before you need it.

You can evaluate physical attractiveness instantly. It's there, it's visible, and you don't need a trial period to find out whether you find someone attractive. The information is available on first contact, and it's reasonably stable over the timescales that matter for the early years of a relationship. Yes, hedonic adaptation will blunt the intensity of your response over time, and yes, bodies age. But the initial assessment is honest. You're not going to be surprised by it.

"Will be with you on the worst day of your life" is a completely different kind of signal. You can't test it. You won't know if it's true until the worst day actually happens, and by then you're already committed. During the dating period, you're observing small-scale behavior, how they act when you're sad, how they handle a bad day at work, and extrapolating from that to catastrophe-level performance. That extrapolation might be right. But it's a guess, not a measurement.

In evolutionary biology, Amotz Zahavi proposed the handicap principle (1975, Journal of Theoretical Biology): reliable signals in mate selection are costly to produce and hard to fake. A peacock's tail is a reliable signal of genetic fitness precisely because it's expensive to grow and maintain. If it were cheap, every peacock would have one regardless of fitness, and it would stop carrying information.

Physical attractiveness fits this model. Developmental stability, symmetric features, clear skin, healthy body composition: these are costly to produce, require genuine underlying fitness, and are hard to manufacture from nothing. You can enhance them at the margin. You cannot fabricate them. The signal is right there, and it's reliable because it's expensive.

"I'll be there on your worst day" is cheap to signal. Anyone can say it, and most people believe it about themselves. Demonstrating it requires a catastrophe, which you can't stage and wouldn't want to. You're selecting on a trait invisible during assessment and verifiable only once it's too late to change your mind, ignoring the high-reliability signal in front of you to bet on one you can't verify.

Loss aversion in a relationship-advice costume

There's a decision-theory angle that explains why the advice feels so intuitively right even though the reasoning behind it doesn't hold up.

Kahneman and Tversky's prospect theory (1979, Econometrica) demonstrated that people systematically overweight losses relative to equivalent gains. Losing $100 hurts roughly twice as much as gaining $100 feels good. This asymmetry is baked into human cognition and shows up in financial decisions, medical choices, and risk evaluation broadly.

The "worst day" framing activates this asymmetry directly. It asks you to imagine the worst-case scenario and select a partner based on that scenario, not what your best Tuesday looks like. That's a loss frame. And in a loss frame, people reliably overweight the probability and the impact of the negative event while underweighting the steady-state positive.

The advice feels wise because it maps onto a cognitive bias that already makes losses feel more important than gains. The advice isn't wisdom. It's loss aversion wearing the costume of wisdom. It tells you that the rational partner choice is the defensive one, the one that minimizes your worst case, the one that functions like insurance. But nobody optimizes their entire life around insurance. You buy insurance and you live your life. The advice conflates the two and pretends this is maturity.

The mutability paradox

Here's the part that actually makes the priority ordering defensible and not just contrarian.

The advice says: pick character over looks. But the traits it calls "character", supportiveness, reliability in crisis, emotional presence during suffering, are behavioral. And behavior is changeable.

The Bleidorn et al. meta-analysis (2022, Psychological Bulletin, synthesizing 189 longitudinal studies with over 178,000 participants) found that personality traits continue to change throughout adult life. Emotional stability, conscientiousness, and agreeableness all show mean-level increases across the lifespan. Rank-order stability, your position relative to others, does increase and stabilizes around r = .94 by age 30, but the absolute levels keep shifting. People become, on average, more emotionally stable and more agreeable as they age. They can also become less so under stress, illness, or changed circumstances.

What this means: the supportive behavior you screened for at age 25 may not predict supportive behavior at age 45. Not because people are unreliable, but because people change. The person you vetted for "will be there on my worst day" might develop in a direction that makes them less available, more avoidant, or simply different from the partner you selected. The screening criteria and the trait's future performance aren't locked together the way the advice assumes.

Physical traits don't have this problem in the same way. Height and bone structure don't change. The fundamental physical characteristics you found attractive at the start remain. Bodies age, and adaptation dulls the response. But the underlying signal is still there. You selected on something real and stable.

The paradox is clean: the advice tells you to prioritize the changeable over the unchangeable, and frames that as the mature choice. But if you're selecting for the long term, stability of the selection criterion is a feature you want, not one you can afford to ignore.

The strongest case against me

I want to steelman the other side honestly, because the contrarian position isn't worth much if it doesn't survive the best counterargument.

The best counterargument comes from attachment theory. Bowlby (1969) and subsequent researchers, Collins and Feeney (2000), among others, established that adult attachment patterns organize around two functions: the safe haven, turning to a partner for comfort and reassurance during distress, and the secure base, using the partner's reliability as a platform for exploration and growth. These functions are among the strongest predictors of relationship quality, individual mental health, and even physical health outcomes. People with secure attachment relationships recover faster from trauma and show reduced PTSD symptom severity.

Here's where it matters: attachment style is more stable than personality traits. It forms early and persists; the concept of "earned security" shows insecure attachment can move toward secure through relationships or therapy, but it changes slowly and with difficulty. Unlike the Big Five, which shift meaningfully across decades, attachment patterns tend to be self-reinforcing. Secure people select secure partners. Insecure people recreate insecure dynamics.

If "will be with you on the worst day" is actually a proxy for "is securely attached," the advice is better than I've been giving it credit for. Secure attachment genuinely is a stable, high-value trait that predicts both crisis behavior and daily behavior. The problem is that the advice never frames it this way. It doesn't say "assess attachment security." It says "pick the one who'll be there on your worst day" — pointing at something real, the safe haven function, through a lens so blurry the practical value drops to near zero.

The other objection worth noting: assortative mating. A meta-analysis of physical attractiveness matching in couples found a mean inter-partner correlation of r = 0.49 (reviewed in Luo, 2017, Social and Personality Psychology Compass), one of the strongest assortative-mating effects in the literature — couples already match on attractiveness roughly within their own range. The advice may be solving a problem that barely exists, people recklessly selecting on looks alone, while creating a new one: people ignoring daily compatibility in favor of a crisis scenario they can't test for.

The ranking, not the binary

I don't think the choice is binary. The advice presents it as binary: the tall one who gives you butterflies, or the one who'll be there on your worst day. Real partner selection doesn't work that way. People are bundles of traits, and you're evaluating the whole bundle, not choosing between two isolated dimensions.

But if you forced me to rank the priorities, I'd put the immutable trait you can verify immediately above the mutable trait you cannot verify until it's too late. Not because physical attraction is more important than supportive behavior. Because selection criteria should be assessed on two dimensions: how reliable is the signal, and how stable is the trait. Physical attractiveness scores higher on both.

I'll admit this advice irritates me in a way that probably isn't purely analytical. There's something patronizing about it. It assumes the person hearing it is too shallow to have already considered that looks aren't everything. It frames the choice as though wanting physical attraction is a failure of maturity rather than a reasonable preference grounded in traits you can actually evaluate. The advice carries a moral charge that it hasn't earned.

The person who makes your Tuesday good is not necessarily the person who performs well in a catastrophe. And the person who performs well in a catastrophe might be terrible company on an average Wednesday. The catastrophes are real, and they matter when they happen. But they constitute maybe a dozen days out of thirty thousand in a long relationship. The other 29,988 days are the ones you're actually going to live.