Join 3,497 readers in helping fund MetaFilter (Hide)


Love Is A Data Field
January 21, 2014 6:22 PM   Subscribe

How a Math Genius Hacked OkCupid to Find True Love
“I think that what I did is just a slightly more algorithmic, large-scale, and machine-learning-based version of what everyone does on the site,” McKinlay says. Everyone tries to create an optimal profile—he just had the data to engineer one.

He went on to write a how-to book: Optimal Cupid: Mastering the Hidden Logic of OkCupid

Amy Webb also tried hacking love.
posted by kyp (101 comments total) 40 users marked this as a favorite

 
TLDR: Lonely computer dude in LA used multiple profiles on OKCupid to get chicks.

Its a movie that pitches itself.
posted by hal_c_on at 6:29 PM on January 21 [3 favorites]


Most were ignored; he’d gone on a total of six first dates.

Six first dates does not foreveralone one make.
posted by ifandonlyif at 6:32 PM on January 21 [4 favorites]


+1 just for the post title.
posted by Greg_Ace at 6:33 PM on January 21 [83 favorites]


The foreword of the book is written by Magic: the Gathering Hall of Famer Jon Finkel, due no doubt to that thing that happened to him.
posted by IAmUnaware at 6:34 PM on January 21 [3 favorites]


This is more about a hacker stealing data than a mathematician with lots of brains.
posted by zscore at 6:36 PM on January 21


Tom N. Haverford?
posted by Roger Dodger at 6:36 PM on January 21 [30 favorites]


So...wait. He realized he was getting a low number of hits because he only answered a handful of the questions. So he went on to...create a bunch of fake profiles and used them to surf other women's profiles to see which questions they answered, so he could then answer those, but then those fake profiles got banned so he consulted a neuroscientist to teach them to act more human, and....

....Um, dumb question, but why didn't just stick with his original profile and just answer more of the freakin' questions?
posted by EmpressCallipygos at 6:37 PM on January 21 [18 favorites]


You missed the part where he optimized the weighting function on the questions he did answer. That was what really jacked up his match %.
posted by zscore at 6:38 PM on January 21 [3 favorites]


Ohhh, wait. So he was trying to increase the quantity of dates more so than the quality?

works for some, I suppose.
posted by EmpressCallipygos at 6:40 PM on January 21 [2 favorites]


The scary part is that even after all that optimization, he went on 55 dates, and then only got 3 second dates, and 1 third. 18:1 first/second ratio, 55:1 first/third.
posted by CheeseDigestsAll at 6:43 PM on January 21 [7 favorites]


Umm I think we might be able to think of this as taking a linear projection of his feature vector (if you consider null responses to be weight 0) in order to maximize the inner product with as many ladies as possible. And EmpressNiceButt, if you have a large quantity of people looking at your profile then you can quickly riffle through them and pick which ones you like best. Prescreening can be pretty helpful, since it saves you time.
posted by zscore at 6:44 PM on January 21


CheeseDigestsAll, he might have been the one turning down a lot of second dates. It seemed like there was something very specific he was looking for.
posted by zscore at 6:45 PM on January 21 [1 favorite]


And EmpressNiceButt,

Maybe off topic, but DifferenceFromTheMeanInStandardDeviations, that came off as a little weird.
posted by Jpfed at 6:50 PM on January 21 [34 favorites]


The article's comments are a mix, but one good observation: Maybe a cubicle isn't the best place for you to live in when you are trying to establish intimacy with a potential partner.
posted by ifandonlyif at 6:59 PM on January 21 [3 favorites]


I recall an SNL skit where they made fun of the dating site, eharmony, basically, answering all those compatibility questions found their perfect match, the seeker's doppelganger; in this case, the female version of himself.
posted by locidot at 7:01 PM on January 21


You missed the part where he optimized the weighting function on the questions he did answer. That was what really jacked up his match %.

But if he'd just answered a bunch of questions to begin with, that would have jacked up his match % too - and would have taken far less time and effort than his approach. Math genius yes, practicality genius not so much.

On the other hand, that wouldn't have gotten him an article in Wired and a book deal, so...
posted by Greg_Ace at 7:08 PM on January 21 [2 favorites]


haha i'm so dying alone

on preview: the female version of himself

At this point as nice as that would be I think the most likely relief would be to somehow disable the circuit that drives this stupid quest for nonloneliness forward.
posted by laconic skeuomorph at 7:10 PM on January 21


And EmpressNiceButt, if you have a large quantity of people looking at your profile then you can quickly riffle through them and pick which ones you like best.

I'm not sure you really need to have so many people looking at you, DifferenceFromTheMeanInStandardDeviations, especially since OKCupid is already set up so you can just look at THEM and riffle through their profiles and pick and choose anyway. Or else, it strikes me that if you want to lure a bunch of people into just looking at you, then maybe you're after something other than what OKCupid is for.

And yeah, calling out my nickname like that is a bit weird.

But if he'd just answered a bunch of questions to begin with, that also would have jacked up his match % too - and would have taken far less time and effort than his approach. Math genius yes, practicality genius not so much.

Exactly.
posted by EmpressCallipygos at 7:12 PM on January 21 [2 favorites]


The true wisdom is near the end.

"But all the math and coding is merely prologue to their story together. The real hacking in a relationship comes after you meet."

No kidding. These folks are getting married. Then the real hacking will begin. It's still a great story and I am happy for them both.
posted by Bella Donna at 7:16 PM on January 21 [3 favorites]


I was expecting to cringe, but I thought he was actually pretty charming. I think being practical about group A was a good call, fuck LA traffic.
posted by kittensofthenight at 7:20 PM on January 21 [2 favorites]


Shrug. After using online dating for nearly ten years, I answered over 1000 questions on OKC, and, after months of bad dates and worse dates, a woman with a 99% match in a town 300 kilometers away sent me a note.

That was two and a half years ago. I moved to her town within six months. We're common-law and own a business and a house together.

Might I have found love earlier if I gamed it algorithmically? Maybe, but I'm pretty happy right now.
posted by ten pounds of inedita at 7:22 PM on January 21 [7 favorites]


I thought this was actually the most effective (and ironically the simplest) hack that he came up with:
He needed one more step to get noticed. OkCupid members are notified when some­one views their pages, so he wrote a new program to visit the pages of his top-rated matches, cycling by age: a thousand 41-year-old women on Monday, another thousand 40-year-old women on Tuesday, looping back through when he reached 27-year-olds two weeks later. Women reciprocated by visiting his profiles, some 400 a day. And messages began to roll in.
posted by kyp at 7:23 PM on January 21 [5 favorites]


Yeah, he could have replied to some questions... or he could have written a Python script! How can you resist?

xkcd_cartoons = [1, 2, 3]
posted by ifandonlyif at 7:25 PM on January 21 [3 favorites]


But if he'd just answered a bunch of questions to begin with, that also would have jacked up his match % too - and would have taken far less time and effort than his approach. Math genius yes, practicality genius not so much.

I doubt that answering a large number of randomly chosen questions would have gotten him the results he was looking for. Dude got 400 women a day to look at his profile(s).
posted by 23skidoo at 7:29 PM on January 21 [2 favorites]


Spending the time relentlessly programming and min-maxing still somehow seems less depressing than actually using OK Cupid to me.
posted by wotsac at 7:33 PM on January 21 [5 favorites]


Which is exactly the opposite of my own reaction, so...different strokes.
posted by Greg_Ace at 7:45 PM on January 21 [5 favorites]


It's cool because if it doesn't work out you can be all "I scripted you stdinto my life, and I can script you stdout!"
posted by lordaych at 7:49 PM on January 21 [13 favorites]


Ohhh, wait. So he was trying to increase the quantity of dates more so than the quality?

"Quantity has a quality all its own."
             -Josef Stalin
                     -Michael Scott

posted by Justinian at 7:54 PM on January 21 [7 favorites]


I doubt that answering a large number of randomly chosen questions would have gotten him the results he was looking for. Dude got 400 women a day to look at his profile(s).


That was the combination of a high match % with the profile-viewing script - when I was on OKCupid, by far the most effective way to get me to look at someone's profile was for them to to look at mine, and show a high match score. But you don't need to studiously game the system to get a high compatibility rating with people - just answer questions, possibly with some effort to not pick the dickish answers.
posted by Holy Zarquon's Singing Fish at 7:55 PM on January 21 [1 favorite]


Or else, it strikes me that if you want to lure a bunch of people into just looking at you, then maybe you're after something other than what OKCupid is for.

Yo OKCupid is like Information Theoretical Semiotics 701 at Hard Knocks University. I'm not saying it's all's fair in love and war but that sounds like a fine tactic to me. Better than "hey girl nice hair" and dick pics.

I myself operate in this sort of thing by constraining myself to true facts and trying to avoid dishonest omission but the many judgment calls, textboxes, and varied but limited options allow for a lot of room to operate under those constraints. Do I drink "rarely" or "sometimes"? List job y/n? Income? (n) Is it possible to meaningfully enumerate 6 entities without which I cannot live, and is it even possible to meaningfully enumerate distinct entities? Do I go with the technically correct/true pedantry answer, or the answer acceptable to the local tribalisms & indeed more superficially indicative of my own views? These varied statistics, options, character progressions, involved dialogue trees, & IRL sociosexual rewards make OKCupid the sleeper browser MMORPG hit of the last decade.

I haven't tried Tinder yet but people are telling me it's like "Computer says you two fuck, y/n?" for straight people, y/n?
posted by save alive nothing that breatheth at 8:03 PM on January 21 [23 favorites]


Math genius yes, science genius no. In going from using-OKCupid-like-everyone-else to his hacking/data clustering/scripting approach, he changed at least half a dozen variables all at once. So it's impossible to say which of the many things he altered was relevant to his success and which was not.

(I'm being a bit facetious—of course his goal was to find a partner for himself, not to conduct a rigorous experiment at How To Succeed In Finding Love On OKCupid. But I think the point stands that his personal experience is insufficient to say what aspects of his approach would be helpful generally, and what would be a waste of time. Which leads me to be skeptical of his book.)
posted by DevilsAdvocate at 8:06 PM on January 21 [2 favorites]


I doubt that answering a large number of randomly chosen questions would have gotten him the results he was looking for

They're not really random, you choose which to answer. If you answer questions about how awesome math is, and ignore all the ones about your astrological sign, you will be doomed to go on a million first dates with grad students.

A lot of people take online dating very seriously and I can't decide if this is more not giving a fuck and gaming the system, or if maybe his seriousness about it is on a different level.

Yo OKCupid is like Information Theoretical Semiotics 701 at Hard Knocks University.

I wonder what our match percentage is.
posted by bradbane at 8:13 PM on January 21 [2 favorites]


I love this story. "Indy" Group A is great as I am probably not young and do have a medium dog. It is so clever I can only laugh. Two dates, same day, same beach. heh!
posted by femmme at 8:16 PM on January 21


I smell a BRo-Com-A-Brewin
posted by lordaych at 8:19 PM on January 21


bong-hitting

guy:dude, you can't use a computer to tell you what love is

guy on computer: oh yeah, i think i just did

--------------...
smiles and kisses montage, pratfall etc
----------later!
her: i never said i could be everything for you
you spent too much time saying people are wrong on the internet
i never got to know you and time is of the essence

guy on computer: oh no, i guess i did int
----------
posted by lordaych at 8:22 PM on January 21


It's way easier to just let your friends set you up with someone.

(It worked for me!)
posted by oddman at 8:25 PM on January 21


Is it possible to meaningfully enumerate 6 entities without which I cannot live, and is it even possible to meaningfully enumerate distinct entities?

Enid: Yeah, yeah, just list your five main interests in order of importance.

Seymour: Uh... I'd have to put traditional jazz, blues and then ragtime at the top of the list...

Enid: Right, so, let's just say music. That way we only use up one.
posted by RobotHero at 8:29 PM on January 21 [3 favorites]


It's way easier to just let your friends set you up with someone.

That presupposes one has friends.
posted by Greg_Ace at 8:37 PM on January 21 [6 favorites]


And that one's friends have other friends. Insular social circles are the bane of dating.
posted by Holy Zarquon's Singing Fish at 8:41 PM on January 21 [4 favorites]


Or that one's friends date outside their ethnic groups. ;-)
posted by viramamunivar at 8:50 PM on January 21


I'm surprised that both the author and his subject so readily accept a distinction between gaming the selection and weighting of questions and gaming the answers to questions. Though I don't know anything in particular about the algorithm OKCupid uses, I'd naively expect weights to be as significant as answers, if not more so.

There's no obvious reason to assume jiggering the weights would be less bad than changing one's answers to generate dates. That's true whether "bad" is an ethical argument against misrepresenting oneself or a practical complaint against making the sorting algorithm work less well by increasing the number of false positives. ("Bad" could even be good, if you believe OKCupid's original scheme genuinely works less well for everyone than the broken system created by tinkering.) Cheating, but only cheating half-way, seems an odd choice. But, I suppose it's only one of several odd choices in the piece.

On a tangent inspired by a user name above, wouldn't it be neat if OKCupid provided higher-order statistics in their "percent match" results, or opened up the matching algorithm itself to explicit tinkering? Letting people invent their own metrics could be really interesting.

What if you could define your own matching metric using as inputs "your answers" "potential match's answers" and the moments of "all answers within a demographic group," especially if questions were arranged into a series of topical dimensions. Then you'd get a list of matches based on your algorithm, and also a score which ranks you according to each potential match's algorithm. eg. Give me a Heaviside step function on two or three major world-view axes, and then take everyone whose answers are correlated with mine above some moderate cutoff and sort them based on how divergent they are from the mean overall.

While you're at it, why not turn it into a game by allowing people to try out other user's algorithms and rate the quality of the resulting dates, with prizes for the most successful. Then, when that gets boring, you open it up for meta-matching based on the similarities of the algorithms themselves, correlating the results of every pair of algorithms when applied to the whole population . . .

I'm also reminded of how disappointing it is OKTrends has died. Despite some questionable data-presentation choices, it was great to see someone digging into what must be an amazing dataset and using it for not-obviously-evil purposes.
posted by eotvos at 8:59 PM on January 21 [8 favorites]


The way the algorithm is reported -- it only looks at the questions you both answered -- seems very simplistic. Wouldn't it make a whole crapton more sense to do some sort of dimension reduction technique over the questions you answer and then plot your locations in however-many dimensions the global algorithm spits out? Then you could match two people who hadn't answered any of their questions in common but had similar ideal points in the romantic hyperspace.
posted by ROU_Xenophobe at 9:22 PM on January 21 [2 favorites]


The questions part is fully cybernetic, eotvos. Your answers are seen from the standpoint of the inner product in the matchmaking algorithm, but also by the other human players for their qualitative evaluation, along with any provided text explanation.

Algorithmically, the effects could be similar, but humanistically, the answers to questions are taken as statements of belief, fact, or honest opinion, with the weighting not particularly apparent (in sidebar as a sort order) and the failure to answer a question considered a signal but not generally a dishonest one.
posted by save alive nothing that breatheth at 9:26 PM on January 21


I'm married to my female doppelganger, just needed luck and stupid life choices to find her.
posted by Divest_Abstraction at 9:28 PM on January 21 [4 favorites]


Am I wrong or does this site, and this thread, completely miss the point of dating?
posted by transient at 9:28 PM on January 21 [1 favorite]


well not all people

They wouldn't fit
posted by ook at 9:33 PM on January 21


Oh nice stealth edit ruining my joke thanks
posted by ook at 9:33 PM on January 21 [2 favorites]


[Reminder, please only use the edit function to change typos, not to add or remove content. Thanks.]
posted by LobsterMitten at 9:35 PM on January 21 [1 favorite]


Sorry; kind of thought better of it at the same time as your comment.
posted by transient at 9:35 PM on January 21


Woe is me a whole quip wasted
posted by ook at 9:38 PM on January 21 [1 favorite]


The way the algorithm is reported -- it only looks at the questions you both answered -- seems very simplistic. Wouldn't it make a whole crapton more sense to do some sort of dimension reduction technique over the questions you answer and then plot your locations in however-many dimensions the global algorithm spits out?

I think it is something like the Peter Principle and 1% knowledge inequality. The people who could just whip that out are making ridiculous radars, NSA quantum crypto shit, motivic cohomology (I am not even qualified to understand the description of what a Fields medal is for, could be Sokal-speak for all I know.), Higgs Bosons, and so on. The people who could sit down and do it workmanlike are engaged as professors, hedge fund quants, and bigger tech fish like Google, Amazon, and Netflix. I'd have a chance but not 100% confidence, but OKC-level web job is like a tier or 2 lower difficulty below my job. Once you get to "Dating website," even a top one, I guess you get an inner product.
posted by save alive nothing that breatheth at 9:39 PM on January 21 [2 favorites]


All I did was use Greasemonkey to inject some JavaScript into EHarmony to add more data to the match list. I had it put profile pictures in the list, heights (not keen on very tall women), distance (complicated script that screen scraped Google Maps directions to the city on their profile page), and possibly some other stuff I can't remember. I still ended up with someone who lived an hour away.
posted by w0mbat at 9:47 PM on January 21 [1 favorite]


an article in Wired and a book deal

It's a self-published 37-page Kindle e-book; less "book deal", more canny self-promotion?
posted by We had a deal, Kyle at 10:00 PM on January 21 [7 favorites]


Man it is creepy and manipulative when people date.
posted by poe at 10:03 PM on January 21


What the hell is the Samantha cluster?
posted by DLWM at 10:04 PM on January 21 [1 favorite]


I think it's named after a character from Sex and the City
posted by RobotHero at 10:07 PM on January 21 [1 favorite]


Ah, that would make sense. I was thinking his algorithms had found some bizarre statistical similarity between people with that name...
posted by DLWM at 10:10 PM on January 21


And that one's friends have other friends. Insular social circles are the bane of dating.

I asked a friend if they could set me up with anyone, and they pointed out that we knew all the same people. This really happens.
posted by yohko at 10:18 PM on January 21 [1 favorite]


The people who could just whip that out are making ridiculous radars, NSA quantum crypto shit, motivic cohomology (I am not even qualified to understand the description of what a Fields medal is for, could be Sokal-speak for all I know.), Higgs Bosons, and so on.

But mostly it's been done already! All you'd need to do is run an item-response model that can handle ordinal input to find people's locations in the n-space, and I know that these exist, and then just compute weighted distances using whatever quantifications go with the weights you assign.
posted by ROU_Xenophobe at 10:21 PM on January 21 [1 favorite]


What I would like to see is a survey of happy long-term couples, and how they would line up using the match-making algorithm. I know that my wife and I probably haven't got more than a 50% match, but we met randomly and really hit it off (and have decided to tolerate or even embrace our differences). Are the scores a useful predictor for long-term happiness?
posted by Harald74 at 10:38 PM on January 21 [2 favorites]


Are the scores a useful predictor for long-term happiness?

They're primarily an indicator of how similar your political and social views are to each other.
posted by MillMan at 11:17 PM on January 21


I'd like to say to his fiancee (who seems really cool and kinda too cool for a guy who did this weird project and slept in his office) in my best Michael Bluth voice re Ann Veal:

"Him?"
posted by discopolo at 11:38 PM on January 21 [1 favorite]


And I hope he didn't claim to be "laid back" or "easygoing" in his profile, because, yuck, he is most definitely not.
posted by discopolo at 11:40 PM on January 21 [2 favorites]


The scary part is that even after all that optimization, he went on 55 dates, and then only got 3 second dates, and 1 third. 18:1 first/second ratio, 55:1 first/third.

He probably should have spent less time going through data and more time "optimizing" his personality so women would want to go out with him again. There are just some guys who have it down and are perfectly delightful and don't need to do all this stuff to find the right woman, and then there are the guys in the Wired article's comment section who want to create a statue of this guy while simultaneously bitching about women being too picky because they don't feel attracted to a guy who sleeps on a dingy mattress and showers at the university gym.

I hope for Tien Wang's sake that the guy does his own laundry and doesn't have to be told that it's time to wash the sheets.
posted by discopolo at 12:01 AM on January 22 [1 favorite]


not sure i have it straight yet, discopolo would you say you are or are not attracted to the subject of the article?
posted by Kwine at 12:35 AM on January 22 [3 favorites]


I think this whole exercise would be really helpful for McKinlay in a meta sort of way. Telling his dates what he went through to find them would be a pretty good acid test to determine compatibility with him,
posted by benbenson at 2:38 AM on January 22 [1 favorite]


It must be really expensive to go on 55 dates, especially if you stick to that old-fashioned idea that men pay for first dates.

If you're eating, then a night out for two can easily be about £40. 55 dates would come to £2,200. And for what?

Fewer, better quality dates seems like the optimal way to go.
posted by winterhill at 2:45 AM on January 22 [2 favorites]


@winterhill, I know a guy who's fit, attractive, stable job with excellent income, friendly, etc. In the last two years, he's been single ~18 months, and in those periods he's gone on about 3 OKC dates a week. They rarely get to a second date. He's not using any algorithms and it's still expensive.

I'm not sure how one determines what the better quality dates would be. McKinlay's efforts were an attempt to find that out, but maybe this is an inherent limitation of online dating.
posted by benbenson at 2:54 AM on January 22


"Everyone tries to create an optimal profile"

No, everyone does not.

Some people do what this guy did, which is basically whatever it takes to get a coffee date. Those people are happy to have the sorting/matching stuff done outside the site, during the meeting. Those people either have a lot of free time, or feel that impossible-to-predict chemistry is the most important thing, or just really like meeting people for coffee dates.

Other people, people like myself, prefer to do all the sorting/matching via the site. You know, use it as designed. I don't "optimize" my profile in the sense that I want to make it most attractive to the most people, I answer profile questions and match questions truthfully and am perfectly happy when some of my answers are "red flags" to many people because — wait for it — I don't want to date those people.

And most people attempt something between those two extremes. Which makes sense and is justifiable.

But the presumption should not be that "everyone" optimizes their profile and is basically presenting themselves like they're some kind of airbrushed model, or that this is a good thing to do. That way lies madness.
posted by Ivan Fyodorovich at 4:22 AM on January 22 [5 favorites]


Most unsuccessful daters confront self-esteem issues. For McKinlay it was worse. He had to question his calculations.

McKinlay's observation, or Wired's?
posted by Obscure Reference at 4:49 AM on January 22


Uh....are we not talking about the fact that after everything, the chick he ended up with found him? She just typed two superficial criteria into the search function, up he popped, and bish bash bosh? I mean, I'm sure all his question jiggering helped a bunch, but when I read that part the wanh-wanh trumpet went off in my brain.
posted by Diablevert at 5:36 AM on January 22 [3 favorites]


Are the scores a useful predictor for long-term happiness?

When I was on OKCupid, I met quite a few guys with over 90% match percentages. None of them worked out in the long run. My husband was something like a 60% match on OKCupid when we met, but in actual real life I can't imagine a better match.

I think the match results must have been comparatively low over relatively superficial stuff. When I was looking, I was in a clubbing, chain-smoking, aimless and cynical time of my life, and I'm sure that affected the answers. I stopped doing a lot of those things eventually, and I'd bet my match percentages would be different today. Not all the way different, but different enough.
posted by Metroid Baby at 5:51 AM on January 22 [1 favorite]


Yeah, to me, the weird bit about online dating mediated by quiz answers is the underlying assumption that everybody is exactly the person they both want to be and will remain.
posted by penduluum at 6:01 AM on January 22


I recently met one of OKCupid's back-end engineers. He describes his job as, "Basically trying to steer guys away from contacting women who are out of their league."
posted by GameDesignerBen at 6:14 AM on January 22 [10 favorites]


I dated a guy I had a 99% match with for eight months or so. It didn't work out. He later told me that he also had a 99% match with his sister.

I'm not sure how much stock - if any - to out in the match percentage.
posted by sockermom at 6:17 AM on January 22


Some people do what this guy did, which is basically whatever it takes to get a coffee date.

I think you're misreading what he did. The actual optimizing he did was to let a machine pick the weights for him, and to emphasize being a math prof in his bio. He at least says that he answered all the questions truthfully -- but he did a lot of work to find out which questions he should answer to maximize his visibility to clusters of women he found interesting.
posted by ROU_Xenophobe at 6:20 AM on January 22 [1 favorite]


Uh....are we not talking about the fact that after everything, the chick he ended up with found him? She just typed two superficial criteria into the search function, up he popped, and bish bash bosh? I mean, I'm sure all his question jiggering helped a bunch, but when I read that part the wanh-wanh trumpet went off in my brain.

Wasn't that...the entire point of this exercise? Configure his profile in such a way as to maximize how frequently he would show up as a high percentage match, thus increasing the potential number of women who would proactively contact him (since his previous experience of sending out messages himself didn't seem to be going anywhere)?
posted by The Gooch at 6:52 AM on January 22 [2 favorites]


Wasn't that...the entire point of this exercise? Configure his profile in such a way as to maximize how frequently he would show up as a high percentage match, thus increasing the potential number of women who would proactively contact him (since his previous experience of sending out messages himself didn't seem to be going anywhere)?

Well, it's a bit of a poTAYto, poTAHto situation -- but as I was reading it he set it up so that he would find better matches, no so that they would find him. He revamped his profile to maximize his comparability with his target population, then attempted to contact every target in order to generate hits --- the reason he was getting all those notes from women is that his profile was crawling the target pop, so that he would come up as a person that viewed their profile. To me, that's him making first contact. Sort of like buying a google ad at the top of the search results. Whereas his fiancée actually found him through, to continue the metaphor, organic search --- she entered what she was looking for and he popped up. Obviously his efforts to increase his visibility to the target pop were a big part of why he popped up for her --- but all the profile crawling still didn't work. At least, it didn't result in him finding that girl --- it did produce 50 or 80-odd first dates.
posted by Diablevert at 7:18 AM on January 22


It's way easier to just let your friends set you up with someone.

It's also way easier to get a job by having your friends recommend you to their employer, but sometimes none of your friends work anywhere that's hiring. I have a lot of very-socially-active friends but it's not an exaggeration to say that about 90% are married or engaged, as are their friends, etc etc. I've been set up a couple of times and it's sometimes worked out for a few dates at least, and in theory I could meet someone through my hobbies, but for the most part, my options are "go online" or "sit at home talking to the cat."

Don't get me wrong, he's adorable.
posted by Tomorrowful at 7:36 AM on January 22 [2 favorites]


It's way easier to just let your friends set you up with someone.

Except when the only single people your friends know are all the same gender, and you're heterosexual. (I'm one of about FIVE single women that one pair of friends I know has been trying to set up for about three years now.)
posted by EmpressCallipygos at 7:44 AM on January 22


He was already sleeping in his cubicle most nights. Now he gave up his apartment entirely and moved into the dingy beige cell, laying a thin mattress across his desk when it was time to sleep.

Did this info go into his profile?
posted by IndigoJones at 8:00 AM on January 22 [5 favorites]


not sure i have it straight yet, discopolo would you say you are or are not attracted to the subject of the article?

I'm head over heels. I'm a sucker for guys with patchy beards.
posted by discopolo at 10:48 AM on January 22 [1 favorite]


It's way easier to just let your friends set you up with someone.

Another problem with this is that friends who like to set you up with people get upset at you if you don't care to date the people they wish for you to date.

OKcupid never bothers you for a play-by-play after you go on a date either.
posted by yohko at 11:50 AM on January 22


@winterhill, I know a guy who's fit, attractive, stable job with excellent income, friendly, etc. In the last two years, he's been single ~18 months, and in those periods he's gone on about 3 OKC dates a week. They rarely get to a second date. He's not using any algorithms and it's still expensive.

I once didn't want to date this one guy anymore just because something about him didnt quite feel like a fit, and after mentioning it to my male friend (who seems to be cool with rejecting women who don't feel like a fit but annoyed when women reject him) who stopped me and said,"I don't get it. He's a nice guy, he's attractive, he works out, he's tall, he has a good job,he likes you, you guys have fun---why don't you want to date him anymore?"

I guess if it were the old days and I was looking for a husband/guy to build a family with, this dude would do in a pinch. But seeings that I'll have to bust my hump working for a bajillion years, I'm going to need more than an average decent guy to get through that with, what with Hugh Jackman being taken and all.

It just strikes me as funny. My friend can say there's something he doesn't like about a woman that's totally superficial (like he didn't want to date this girl who enjoyed fashion and apparently women into fashion are shallow or something) but it's doubtful he would have settled for good job, stable income, nice enough girl. He wants a traditional girl who still looks fantastic but it has to look effortless. Yet I'm supposed to settle? And then these guys bitch about how women they want being too picky. I get that no one likes being rejected, but if women have to nut up about it, guys really should too without getting bitter. And maybe spend more time thinking about how to be better people and partners.

The guys who have the hardest time on OKC are probably the ones who need to learn to treat women like people. My gut tells me that Tien Wang is in for deep disappointment if she actually marries this guy. He doesn't seem like the kind of guy who works on improving himself or his ability to relate to women; he seems like the kind of guy who expects women to adapt to him.
posted by discopolo at 12:14 PM on January 22 [3 favorites]


As long as the "real hacking" after marriage doesn't involve axes.
posted by gottabefunky at 12:16 PM on January 22


This is all about the math and the clever tricks he used, and I'm sure that helped. But I also strongly suspect that his dating success was more a matter of trial and error, practice, and basically learning how to date better.
He developed a set of personal rules to get through his mara­thon love search. No more drinking, for one. End the date when it’s over, don’t let it trail off. And no concerts or movies. “Nothing where your attention is directed at a third object instead of each other,” he says. “It’s inefficient.”
It seems obvious to me that there is a significant "socialization" component to this story that Wired is glossing over in favor of details about the bot programming and the algorithms he developed.
posted by ErikaB at 12:24 PM on January 22 [4 favorites]


(Don't get me wrong, I think this kind of clever data optimization is really hot, and that's not sarcasm. But I think there's more to this story than he realizes.)
posted by ErikaB at 12:26 PM on January 22 [2 favorites]


This is brilliant. Thanks for posting.
Should he repeat the experiment to verify his results?

Round here the geeks just learn how to Lindy Hop, a safe community and dozens of 4 minute hug based micro-dates in one night that tell you most of everything you need to know about someone.
That's quicker optimising of a search algorithm than any cubicle maths.
And a few years later... I'm left here looking after the baby tonight while my wife goes out dancing ;)
posted by Dr Ew at 12:28 PM on January 22 [1 favorite]


Wow, there really is an XKCD for everything.
posted by EmpressCallipygos at 12:30 PM on January 22 [1 favorite]


With regard to OKC's matching numbers...

People should be aware that OKC matching doesn't find a duplicate of oneself. All the match questions ask both for your answer, and how you want your ideal partner to answer. So matching is each person's own answers against the other person's ideal partner answers. Depending upon how people answer their question, a high match percent could be someone very unlike oneself.

This doesn't avoid the problem of people not really knowing what they want in a partner, however.

The vast majority of potential partners are very much nothing like people I'd want to date. I look at photos after I've checked the match numbers and the profile answers because at my age and where I live, the majority of potential partners are people I couldn't even have a conversation with, mostly. For the entire KC metro area, I match greater than 90% with less than ten women in my age range.
posted by Ivan Fyodorovich at 12:30 PM on January 22


I kind of love this, and it's just a more rigorous version of what I've done with my profile (hetero female in her 30s). I've stopped answering the questions I don't care about on the theory that they won't improve my matches. All he's doing is tailoring which questions he answers based on what he expects the women he's interested in to care about (NOT the actual answers). To me, that's no different than, say, going to a bar that caters to the type of people you enjoy being around. I don't see anything dishonest or icky or objectifying about that. I would totally date this guy.
posted by natabat at 12:52 PM on January 22


Did the mathematician who hacked OKCupid violate federal computer laws?
posted by homunculus at 2:36 PM on January 22


Almost certainly. If you do anything remotely dishonest on the internet, you're a felon.
posted by Holy Zarquon's Singing Fish at 5:00 PM on January 22


And someday soon you may not even be able to talk about it.
posted by homunculus at 5:15 PM on January 22


OK the thing I am really interested in here is the clustering of the women's answers. And I would freakin' love to generate those for m4m, w4w, m4w as well. Too bad I am way, way lazier than the dude in the article.
posted by en forme de poire at 6:08 PM on January 22 [1 favorite]


What surprises me most is that OkCupid doesn't have some anti-script tactics. I'm pretty sure they don't like that non-humans use scripts to inflate their back-visits like that.
posted by ymgve at 10:08 PM on January 22


Interesting to see how many people go on a total of 60+ dates and find their true love anyway without all this 'optimisation'. How far ahead of the curve did he really get?
posted by obiwanwasabi at 11:07 PM on January 22


What surprises me most is that OkCupid doesn't have some anti-script tactics.

You missed the part where his first round of bots all got banned so he programmed the next batch to match a real person's click rate and typing speed.

Which if he can generalize that outside of OKC there have got to be customers out there for that sort of software I'm thinking
posted by ook at 5:44 AM on January 23 [1 favorite]


Sorry, Guy: I Won’t Come Work For Buzzfeed
Last week Wired magazine published a feature article on how I hacked the dating website OkCupid and met my fiancée. I subsequently received the following job offer via email from Buzzfeed’s Director of Partnership Development:
posted by the man of twists and turns at 7:07 PM on February 1


Love Is A Data Field

When Big Data Meets Porn: A new tool lets you chart people's sexual searches.
posted by homunculus at 8:03 PM on February 1


thanks but no thanks Buzzfeed. Why don’t you ask Katie Heaney? She seems to know a lot about probability and besides, she’s your perfect match

oooh feel the burn
posted by ook at 9:49 AM on February 3


« Older Grindcore fan that happened to be a goat passes aw...  |  Old photos of the Cincinnati P... Newer »


This thread has been archived and is closed to new comments