Dating: a Research Journal, Part 1

Part 1 – The Best of You

“There is always some madness in love. But there is also always some reason in madness.” – Friedrich Nietzsche

Love is a capricious spark, a miraculous whirlwind. It is found by following ancient prophecies, embarking on dangerous quests, facing your deepest fears, and finally realizing it was there with you all along. Something like that, who knows. Anyway, it sounds like finding a girlfriend was crazy hard before computers.

No more! We have game theory and algorithms, we have OkCupid and Excel, and we’re going to optimize the shit out of happily ever after. I am not claiming that love is reducible to numbers-crunching, just that data and math are better guides than most “conventional wisdom” on the subject. I am deeply in love with my girlfriend, my feelings for her have nothing to do with the fact that a decision matrix may have been involved along the way.

The scientific method

“Scientists are human—they’re as biased as any other group. But they do have one great advantage in that science is a self-correcting process.― Cyril Ponnamperuma

This is written as a guide, but it is really just my own story. Optimize thyself, beware of optimizing others. I will write exclusively about approaches that I have tried myself and found to work, I’m not going to spout a generic advice column. On the minus side, I can not tell you how to apply the lessons to your own situation.

Here’s my situation: straight guy in my late twenties living in NYC surrounded by an endless dating pool of women my age. I am looking for long-term relationships, but I also cherish friends and sex (in that order). I am picking only for myself, there is no one else (parents, pastors) that I need to satisfy. I have passable looks and charm, but I am better at writing than I am at dancing. If you don’t like my writing, that tells you all you need to know about my dancing.

Worse than this

In the summer of 2014 my girlfriend of four years and I broke up and I signed up for OkCupid. Around the exact same time (correlation doesn’t imply causation!) I discovered LessWrong and the idea that I can use my brain to optimize things. I rejected the viewpoint that relationships “just happen” and treated dating as a skill that I can improve with practice. In the first couple of months I went on three or four mediocre dates. I rewrote my profile better and weirder, changed my approach to finding matches, practiced being my honest but best self on dates. I read OkTrends and Dataclysm, twice. I also got a sense for women who would bore me to death 10 minutes into a date and the ones who can inspire me. I used spreadsheets.

As the months passed by, I enjoyed some transient flings, made friends that I’m still in touch with and found a travel buddy that went to Brasil with me. I rewrote my profile again and changed the photos. I started meeting more amazing women more often: a fashion designer, an actress, a civil engineer, an investment banker, a comedienne, three med-school students and a cheese maker. I dated an exquisite 6’1″ computer scientist who dumped me when she got an academic fellowship and wanted “to stop dating for a while to focus on AI research”. I dated a cool journalist with a unique approach to OkCupid who wrote about it in the Washington Post. Finally, nine months ago I met not-her-real-name Rachel. She’s stunningly beautiful, but I didn’t feel struck by lightning on the spot. It took us a few weeks to really get to know each other, but by then I knew exactly what to look for, and I found all of it and more.

These are my credentials, here’s how it happened.

I braved the interwebs

“OkCupid, the vast, weird pink-and-blue toned jungle of the id masquerading as a dating site, where rare birds of modern romance flutter amongst the night-terrors of human loneliness and despair and the suspicious skin irritants of late-night hook-uppery.” – Laurie Penny

comp advantage
Ricardo’s illustration of comparative advantage.

Like most guides about dating, we begin with a review of an 1817 book on the macroeconomics of international trade. It outlines the principle of comparative advantage: when trade and substitution are possible, each country (or company, or person) should do what they are best at, even if someone else is better at it. In the classical example Portugal more efficient than England at producing both wine and cloth but both countries benefit from specialization. Instead of spending 170 hours (England would spend 220) making 1 unit each of wine and cloth, Portugal can work 160 hours to produce 2 wines, England works 200 hours to produce 2 cloths, and they can trade one of each. Each country ends up with the same outcome (1 wine, 1 cloth) and they gained an extra 30 hours to chillax or produce something else.

In the seduction market, people bring different skill sets. Someone has great moves and picks up ladies on the dance floor, someone lights up Tinder by having great abs and not much to say, someone writes well and succeeds on OkCupid. The point of comparative advantage is that however you compare to others, focus on what you’re best at. The writer “trades” the opportunity to meet girls at clubs to the dancer by staying away, the dancer has enough club-ladies to himself and he leaves OkCupid to the writer. If your comparative advantage is growing corn, awaits you. If your main asset is a defect in the melanocortin 1 receptor protein, will allow you to perpetuate your mutation in the population gene pool without fear of competition.

If your comparative advantage is maritime navigation, you’ve got

Comparative advantage plays in planning the actual dates as well: I am at my best in high-engagement conversations, cracking jokes and diving into serious topics. On first dates I would never go to a show, a movie, or even a loud bar, any place that precluded easy conversation.

Even if you’re not a wizard with words, dating sites also offer an absolute advantage: the sheer number of users. How can you find love that’s one in a million by hanging out at a bar that welcomes at most five new faces every night? Do “friends of friends” introduce you to 20 potential dates a day?

I treated it like a full time job

“Love isn’t something you find. Love is something that finds you.” Loretta Young

“And where did this lazy, self-indulgent attitude get you, Loretta? Seems like it got you four solid decades of scandals and divorces, in spite of your beauty.” – me

“Finding love is like making creme brulee. It may take a few tries before you get it right.” Crystal Woods

“Now that sounds much more sensible.” – me

Why are people seduced by the pernicious meme that finding love requires no deliberate effort? Some people stumble into great jobs via friends or lucky breaks, but surely no one will advice a job-seekers to “relax and let it happen”. You can’t predict exactly when labor will turn into lovers, but if you don’t work at it you keep your chances are slim.

I work 50 hours at my job, if dating was at least 10% as important as my career there was no excuse to spend less than 5 hours a week on it. My routine was carving out a 2 hour block (8-10 pm are statistically the best times) to work through matches. On average, I would see around 300 profiles, click on 60, write 20 personal messages, hear back from 5, develop good conversations with one or two and set up the dates. I did this every week, diligently and methodically.

Besides just getting to date a lot of lovely ladies, the methodical two hour sessions allowed me to look at my process systematically. I saw which kinds of messages were getting traction and which profiles led to interesting conversations. This allowed me to target more accurately and message more effectively.

Engaging with a lot of people also gave me a good idea of how my dating pool looked like. The most shocking thing I discovered was that “out of my league” didn’t exist. I got a similar reply rate from the best looking women and the average ones.Replies-Attractiveness-Male-Sender.png

The boring way to read this chart from OkTrends is that looks matter: handsome guys get more replies, pretty girls reply less often. Another way to read it is that the schlubbiest guys get a 12% reply rate from the hottest women. 12%! It’s almost like women seek other qualities in men besides their looks, unbelievable! No matter how insecure you are about your appearance, you need to message only 8 women on average (the number has gone up a bit since 2009 but not dramatically) to get a response from a hottie.

Rachel and I had a chance to chat with one of OkCupid’s passionate developers, she showed us data on message statistics. Based on the response rate both of us were averaging, we calculated that she is objectively 2.2 times better looking than I am (she loves me solely for my skill in Bananagrams). How did I get her to go out with me? I messaged first, and I mentioned a microscope.

I got my message out there

“If you want a love message to be heard, it has got to be sent out.” – Mother Teresa

Ladies and exceptionally hot gentlemen: you have to keep writing messages even if your inbox is full. The reason is glaringly obvious to everyone who studied algorithms and is familiar with Gale and Shapley’s solution to the stable matching problem, which I assume is true of most OkCupid users.

To those few who are still in the dark: the classic matching problem example deals with two equally sized groups who are looking to pair up with each other, like men and women. Each person has a preference ranking over the other group: each dame and dude orders their prospective partners from most to least appealing. A stable matching is an assignment of pairs such that there isn’t a guy and a girl who both prefer each other to their assigned partners. For example, if Alex is matched with Bonnie and Clyde is matched with Dora, but Bonnie and Clyde like each other more than they like Alex and Dora, the matching isn’t stable because Bonnie and Clyde will break the arrangement and go on a crime spree. We don’t want that.

The solution is pretty straightforward: one group get to be the “proposers”, one group the “choosers” (in a clear case of bigotry, a solution isn’t guaranteed to exist for a single group looking to disperse into stable pairs, e.g. gay men or women). If the guys are “proposers”, in the first round each guy proposes to the girl he likes the most and each girl gives her favorite among the proposals a “maybe” and the rest of them a “no”. In each subsequent round, guys who got a “no” move on to their second-choice ladies, then their third choice, and so on. After a certain number of rounds, each girl ends up with a single suitor she kept as a “maybe”. At that point the “maybe” becomes “fine, I guess” and that’s the final arrangement.

A schematic illustration of the Gale-Shapley algorithm

It seems like girls have it easy picking and choosing when guys have to do all the work, but there’s an interesting caveat to the result. An old mathematics professor explained it to me when I was 14 and it changed my life forever. Each person in the “proposers” group is guaranteed their highest ranked match of all the ones they could possibly end up with as part of a global stable arrangement. Each person in the “choosers” group, women in our example, is guaranteed their worst possible match.

This arcane theoretical result applies in full force to the real OkCupid data. Most messages on the site are aimed at recipients who are much more attractive than the senders. Your inbox should be a gauge of the absolute least you should be willing to settle for. Pick the best person in your inbox and write 10 good messages to 10 people who appeal to you more, you’re almost certain to get a reply. This applies especially to women messaging guys, as ladies get a higher response rate at each level of looks. Even the women judged by male users to be unattractive get an average 25% response rate, I think it’s easier to write four messages than to cast a spell that makes your breasts bigger.



Messaging is work and people are lazy, use this to your advantage. Rachel didn’t think she ever needed to message guys first: her inbox was overflowing with suitors! And now, she’s stuck with me. Let this be a cautionary tale.

   Ever wrote
   Got ignored
   No matter

   Write again
   Write some more
   Write better

OkTrends has some basic tips for writing a good first message: be literate, choose self-effacement over direct compliments, and keep it short especially as you’re going to message a lot of matches:

A lot of the shorter messages are of the ‘hey u sexi’ variety, reply rates for literate letters don’t increase much beyond 500 characters (two sentences).

To these I will add a crucial tip: make it obviously apparent that you aren’t copy-pasting. In Dataclysm, Christian Rudder finds that a huge number of messages on OkCupid are mass copy-pasted spam (he figured this by analyzing keystrokes: if it took you two keys to write 50 words, those keys were Ctrl+V). These spammers aren’t only gems like ‘u R hot want 2 cum ovR’, but also generic missives like ‘Hi, I read your profile and I find you very interesting. I think we should get to know each other.’ Since mass-spamming is so quick and easy, it only takes a few spammers to fill every girl’s inbox. Experienced users sniff these out quickly. I strongly urge against being a spammer yourself: most of your time should be spent searching for the best potential matches. Once you find a great potential partner you should take the time to write them the best message you can come up with.

OkTrends recommends writing about shared interests, I would go a step further: your first message must be about a unique topic that your addressee mentions on their profile. This gets the conversation going by letting them talk about their interests, and just as importantly it proves that you read their profile and took the time to craft a personal message. The communication doesn’t need to be convoluted, people will check out and judge your profile anyway before replying. The note must be personal and encourage conversation.

I like to go for the most unusual topic on someone’s profile/photos (e.g. the microscope) – let a girl geek out! If a girl likes both restaurants (as do 99% of New Yorkers) and paleontology (as do 1%), I’ll ask if she saw the new Titanosaur at AMNH. If she has recent photos from a trip to Japan I’ll ask what places she recommends to visit in Tokyo. If her geekiest interests match mine, bingo:

That’s the one who’s programming a pole dancing AI

I stood the fuck out

“Hold on, is your profile picture really you hanging upside down from a stripper pole in your underwear? This is a joke, right?” – Everyone just now

This is certainly not a joke. This is a deliberate decision made after weeks of research and testing. More than anything else I did, this picture maximized the amount of ambitious, intelligent women who wanted to talk to me. All I wrote so far in this post you could have figured out on your own, this is the super-weapon.

First, the data. The most important post on OkTrends explains that women get a lot more messages if guys differ wildly in assessing their appearance, i.e. if some guys think they’re super hot and some guys think they’re unseemly, with no in-between. I looked at speed dating data from Columbia University and confirmed the effect for both men and women. When controlling for average rating of attractiveness, the variance in ratings correlated strongly with more people wanting to date you.

Regression of messages received on ratings of attractiveness

Christian Rudder found that in fact each person rating you a 4 out of 5 correlates negatively with your popularity. Ratings of 5/5 have a large positive effect of course, but so do ratings of 1/5! Rudder explains that people hesitate to write “conventionally pretty” girls (those who get a ton of 4s and some 5s) but when they see a 1-or-5 girl they think “she’s all mine!”

My theory is slightly different: people have idiosyncratic preferences and specific tastes. When you see someone who fits your “type” (even if you didn’t have a rigid type in mind), you want to give them a 12 out of 5 and fill their inbox with love letters. Maybe most people prefer the casually pretty woman on the left, but those that are into the punk rock chick on the right are really into her. The opposite of love isn’t hate, it’s indifference and rating someone 3/5.


A singular look gets noticed and remembered. “Browse matches” on OkCupid is an infinite scrolling wall of thumbnails, and so are the inboxes of the worthwhile sweethearts. If you don’t stand out, you don’t exist. When you’re hanging upside down from a stripper pole in your underwear and a bright pink shirt, you stand out.

Being weird achieves another crucial goal: filtering. Finding love is the opposite of going for mass appeal, you’re targeting a niche market of one person (or five). The more exceptional your profile is, the better matched will be the few who dig it and the deeper they’ll dig it.

When I noticed that my popularity increased, I ratcheted it back down by making my profile more unusual and more honest. I peppered in dumb jokes (“dating me has been shown to help you maintain healthy skin and protects your phone from viruses”). I mentioned being a math geek and studying rationality, with multiple links. I wrote a list of nine qualities I look for that included “altruistic” and “slutty” (but mentioned that I can work with 5/9). My profile was too long, too vague, too self-indulgent. The first lines on it were a silly rhyme:

   Brand Differentiation – A Poem

   Men post photographs, half-naked
   And demand you get to know ’em
   Market’s over saturated
   To stand out, I wrote this poem

There are 729,500 single women my age in New York City. My picture and profile successfully filtered out 729,499 of them and left me with the one I was looking for.

This concludes Part 1, which focused on ways to improve your own success at starting conversations with potential dates: work hard and methodically, ignore your inbox, write personal messages, stand out, stand out, stand out.


Part 1.5 is an advanced technique on OkCupid for increasing your visible match percentage with all the relevant people.

49 thoughts on “Dating: a Research Journal, Part 1

  1. I wish you had mentioned you were doing this in New York in your first few paragraphs. Demographics, demographics, demographics.


    1. Um, it’s in the fourth paragraph?

      I agree that NYC demographics are different from most places, but I think that the differences are vastly overstated. People usually focus on three things:

      1. 8 million people
      2. Slightly more women
      3. Many liberal, secular and educated people.

      1 isn’t relevant beyond a certain point, I am sure that I haven’t seen even 1% of the profiles of women my age in NYC. As long as you don’t live in a tiny, remote village you’re OK.

      2 isn’t that important online, it’s not like a small party where a 10-9 gender imbalance makes a huge difference in the dynamics. Every good looking girl gets dozens of messages anyway, and I think the popular idea that guys can afford to “fool around” because of the gender ratio is a bit overblown. First of all, most guys are in guy-heavy careers like finance and vice-versa, so gender imbalance isn’t felt that much in day to day life. Also, I think that the interpersonal differences in dating goals and desire to commit are much bigger than the effects of the city you live in.

      3 is quite important, but no matter what your religion / social class / education you probably live in a place that has other people like you. If you don’t, this basically reduces to point #1: it may be hard to find an atheist partner in Salt Lake City but you also face less competition. As long as there are at least a few thousand potential matches, you should be good. And if there aren’t, I feel for you.


        1. Correction: I no longer think this is remotely good advice in NYC. The demographics are better enough that you no longer never get responses no matter what, but non-top-10% guys are still going to get functionally no responses to anything other than full right swipe spam.


  2. Apologize if you answered this, but what happens after you have lots of first dates? Do you make further plans whenever possible, do you mention that you’re seeing other people? What if more than one of them likes you? What if you go for one person and then it doesn’t work out, do you try another person you hadn’t yet followed through on?


  3. Most messages on the site are aimed at recipients who are much more attractive than the senders. Your inbox should be a gauge of the absolute least you should be willing to settle for. Pick the best person in your inbox and write 10 good messages to 10 people who appeal to you more, you’re almost certain to get a reply

    I’m willing to give it a shot, but I’m skeptical. I’ve been messaged by some very attractive persons on OKC and some more … shall we say short term oriented venues. The catch is that they never follow through after a few messages (its possible that my writing is the problem, but I doubt it – less attractive persons are usually willing to follow through – I’ll ponder this some more).


  4. I agree more with your theory rather than the OKtrends game theory explanation of the variance of attractiveness vs. messages received. I think it has less to do with the perception of less competition and more with signaling a good catch.

    I think some people are generically attractive and some people are uniquely attractive. There’s more consensus on the former and more variance on the latter. There’s plenty of “barbies” out there who are good looking to most guys most of the time but generally speaking people prefer authentic partners not clones. They want what others don’t have which is the uniquely attractive person. Signaling a good catch might not be as important in a small population/lower competition environment.

    I think this explains popularity in high school. The most popular kids were uniquely popular, just trendy enough without “trying too hard”. Whereas, the block of people right under them (still the second most popular group) just copied their style and emulated them poorly en masse giving the impression that they were “working” to be popular. The second group wants what they can’t have and so it just reinforces the impression of a king/queen/leader relationship to their legion of fans.


  5. Great post, as per usual. Small grammar error in this sentence:
    “In the classical example Portugal more efficient than England at producing both wine and cloth but both countries benefit from specialization.”


  6. “the schlubbiest guys get a 12% reply rate from the hottest women.”

    Yeah, but what percentage of those replies are: polite rejections, insulting rejections, or even bots?

    Liked by 1 person

Leave a Reply to Jacob Falkovich Cancel reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s