What’s Wrong With the OkCupid Algorithm for Matching People?


My blog assistant sent me this link about the science behind OkCupid’s algorithm.

I watched the seven minute video and while I found it entertaining, I saw two big flaws in the way OkCupid purports to match you based on compatibility:

1) OkCupid has three pillars of their algorithm – what you think, what you want your partner to think, and how important it is to you.

This all makes sense in theory, but it falls apart in practice for this reason: people don’t know what’s good for them. Really. If they did, I would be out of a job.

Simply put: what you’re attracted to and who you’re compatible with are two very different people. Most people – myself included – are wildly attracted to the opposite sex versions of themselves. Do you like skiing? Then HE better like skiing!? You’re charismatic? HE better be charismatic! You make a lot of money? HE better make a lot of money! Such commonalities may cause you to be more attracted to someone, but none of these things are good predictors of true compatibility.

That second pillar of the OkCupid algorithm doesn’t say much of anything – it only goes to show how narcissistic you are in searching for someone who is just like you.

As such, that second pillar of the OkCupid algorithm doesn’t say much of anything – it only goes to show how narcissistic you are in searching for someone who is just like you. For this algorithm to work, people would have to be a LOT more self-aware about compatibility – and they’re not.

Because you could find a neat-freak who likes being the center of attention, but if he doesn’t believe in marriage, or doesn’t make a livable wage, or doesn’t believe in sharing household duties, it doesn’t matter.

2) The way OkCupid weights importance is seriously out of whack (see 4:43 in the video).

Irrelevant = 0
A little important = 1
Somewhat important = 10
Very important = 50
Mandatory = 250

A little important is BARELY scored higher than irrelevant. Mandatory is TWENTY FIVE times more important than “somewhat important”. I’m positive some very bright people put some thought into this and decided on this scale, however I don’t know how “somewhat important” can be ten times more valuable than “a little important”, while “very important” is only five times more valuable than “somewhat important”. It’s all very arbitrary and inconsistent, if you ask me.

Frankly, I think eHarmony’s compatibility testing is probably smarter and more relationship-based than OkCupid. The problem, of course, is that eHarmony took all the fun out of their questions and removed attraction from the equation. So you have OkCupid, which is fluffy compatibility pseudoscience based on physical attraction and common interests (as if mutual love of horror films has anything to do with anything) and eHarmony, which is undoubtedly deeper, but ignores the base human impulse to look up attractive faces. Most people I know are using OkCupid because it’s more fun, not because it produces better matches.

Your thoughts, below, are appreciated.

Join our conversation (62 Comments).
Click Here To Leave Your Comment Below.


  1. 41

    Not sure if this was addressed, since I didn’t read through all the comments, but I find a number of other flaws in the algorithm.   Say one person answers 6 questions and another person answers 600 questions, Okcupid will still show a compatibility  score for both prospects.    Yet how is it possible with so little information for the tally to be accurate in showing compatibility?    And what if one person answers a  series of  banal questions such a favorite ice cream flavor and another person answers questions citing core values,  again, their compatibility percentage will still be visible.   A date I was on suggested Okcupid also cite a reliability percentage but I think a better solution would be to not have a compatibility score visible unless both prospects answer a variety of questions across  a spectrum.   The spectrum being questions having to do with personality, values, interests, etc.   And questions such as, do you want to have children should carry more weight (scored differently) than questions lower in importance  which  are not necessarily deal breakers.  

  2. 42

    I don’t know how scientific it is, but I met my boyfriend on OKCupid and he’s about as perfect for me as he can get.   lol   Maybe I just got lucky?

  3. 43

    I found the questions on OKC were funny and helped me learn something about each individual that I was curious about. I’m newly widowed (married for 22 years). So the whole online dating world was very new to me. My son recommended the site to me because he met his current girlfriend on the site too. They just celebrated nine months together.
    I set up my OKC account on a Thursday and shut it down by Saturday because I had found a great match in a man that believed and wanted the same things as I do.
    The nice thing about the account is that I was immediately contacted by a dozen men of all of different nationalities and ages. I’m bi-racial (Filipino/black). I’m a BBW which didn’t stop any of them from contacting me. Whatever I wrote must of been good because of the prospects that contacted me. They were all pretty much serious inquiries. I weeded through them and found the one that I am currently talking to now.
    I’m now in a committed long distance relationship and am very satisfied with the direction our relationship is going. We take each day in stride. He calls me every morning and evening. We constantly text. We talk about our future together. I’m very excited to see the hope and joy that this new man brings to my life and now to my future.
    If I hadn’t had such good luck my first time around looking for wonderful man, I would of seriously considered hiring EMK. But the information provided on his site and the emails that I’ve received are invaluable and are helpful to me in my current relationship. I will continue to read his posts and the insightful comments provided by those who post too.

  4. 44

    With online dating, people have the experience of meeting someone in person after the online ritual of viewing profiles, emailing, calling, and fantasizing about the profile/photo.
    I went to   a Match stir event that turned this experience on its head. I talked to several women throughout the night, and when I came home to view their profiles, I was stunned to find out they were nothing like their profiles. They were much more interesting and attractive in person than their profiles.
    There were also men there that were the most socially awkward people you could meet, but when I viewed their profile, they portrayed themselves as the life of the party!
    I thought “how could this be?”. Is online dating that fundamentally flawed as far as gauging chemistry, attraction, and how a person really appeals to you?
    My take away from this event is that all of this talk about online attraction and algorithms is really moot when you comes to meeting someone. In fact, sifting through profiles is such a flawed way to find a mate because we’re weeding out people who we could be attracted to in real life . . and spending countless hours, days, and weeks emailing people to only find out when we meet them in person after 5 secs that there isn’t a mutual attraction there.

    1. 44.1

      I have spent about a year on OKC both in the US and briefly in Europe. My experience has been identical so I will lump them together.   We may argue about the weighting of the match questions (as EMK does) but I found them to be workable.   However, only from a certain number of questions answered on up.   I answered a total of just under 2000 questions; mostly reply to questions answered by profiles I was interested in (sometimes just generally curious).   Unless the answers come solely from one topic, at around 100 jointly answered questions they become solid. From 200 jointly answered questions, they are 99% stable.   I matched my own ratings against those of two former GFs and one regular friend found  them remarkably accurate.   There are outliers but not so  many.   I still think  OKC is no better than any other site  because in m experience, a good match score means nothing.    Even a 96 match, 54 friend, 03 enemy match will not gain  you even a reply, much less a date on their own.   As it is everywhere else, it comes  down to  only two factors: Picture and age. I have received mails from women  with a 51 match, 40 friend, 35 enemy match and similar because, I assume, they found me attractive, but being 42 (or over 40) makes you invisible to most every  women young than 36 regardless.   Lastly, as it is written elsewhere, it is a numbers game.   There’s truth to the ‘having to  kiss a lot of frogs to meet your prince’ but in the day and age of online dating, where the boutique mentality prevails and the idea that perfection is attainable and waits likely just around the corner, women are  even less likely to kiss any frog or even have coffee  with him lest she sees “potential”.   In the picture. And the age.   🙂  

  5. 45

    Im an active Okcupid user since April 2013 and i still havent found my match. Sometimes the site suggest me people as “matches” and when i look into their answers, i dont see how we would get along, some of these guys are even assholes who dont know how to treat a women, judging for their answers to the questions so i keep using the site and i will keep using it but now i wouldnt bother with the match, i would go right to their essay and then their answers to the questions, thats only how i would know if he a potencial match, i dont care that much anymore if we have mutual interests, just that he is mature, fun and knows how to treat a lady.

  6. 46

    OKCupid provides a lot more  information about people  than any other website —  thousands of multiple  choice questions, created by users,  that can also be commented on.
    The questions cover every conceivable thing that could be asked about.   To the extent that people answer them, you can figure out a lot about  where they’re coming from.
    What more could you reasonably ask for (?)  How well you use it  is up to you.
    It’s also multi-leveled — since the questions are created by users,    there are subtleties and nuances built into some of  them  (if you can tune into that).
    For example: there’s a question (that almost everyone answers) about which is bigger, the Earth or the Sun.   What would you answer (?)

  7. 47

    I just started with the online dating world this past weekend and OKCupid’s software chastised me for marking so many things irrelevant.   So I really tried to have a preference about a guy’s answers to these inane details and then I had a useful revelation.   I’d think about the last two men I fell in love with and notice that we rarely had these details in common.   What made those relationships work was both men had their own interests and were totally secure with my running off to do my thing with other people, so I got to learn some new stuff from them, and we had lots to talk about when we were together.     And then when I bumped into this site, one of the first posts I read was how women are attracted to men who combine passion and competence, so now I’m reading lots of the blog ’cause EMK’s insights ring true.

  8. 48

    OKCUPID = OKSTUPID! Ive been on and off this site for over a year and have NOT gone out with one decent guy….They have all been incompatible! I like POF better! The people on this site are NOT relationship minded and have given out my phone number to some and they don’t even call lol….Not looking for a pen pal thx! I think these guys on here are lazy avoiders that give these sites a bad name…the just wanna hit it and quit it…I actually deactivated my account a week ago cuz i got so frustrated.

    1. 48.1
      Karmic Equation

      Just like IRL, there are good guys and bad guys on dating sites.
      You have to exercise good judgment in discerning one from the other.
      That said, recently, one guy seemed really cool. We exchanged #s. His texts started out ok. Then he got a little weird, so I had to block him. First one from OKC. I had that happen with one guy   on POF.
      I’ve probably corresponded with well over 100 men and had first dates with a good 30 of them. I’ve always treated dating as a fun night out making a new friend and flirting if he’s cute and being a little more serious if he’s not so cute. 2nd dates almost always happen with the cute ones, whereas it doesn’t with the ones with whom I’m less than flirty. I just can’t bring myself to flirt with someone who isn’t cute to me.
      If you’re putting pressure on yourself to find “the one” on each date, that’s when you’re going to get frustrated.
      I look upon each date as honing my flirting and communication skills. It’s a lot of fun to get guys to talk about themselves so you can get a better sense of who they are and what interests them. And then you figure out if that stuff is compatible or not with your own interests.

  9. 49

    My husband and I met on OK cupid. What does that say about us? I guess according to your logic we are both “narcissistic”.

    1. 49.1
      Evan Marc Katz

      Just because OkCupid’s algorithm measures how much “alike” people are based on silly common interests (a method of choosing a partner which I termed narcissistic) does NOT mean that two people who both like running are inherently narcissistic. If you conflate my critique of OkCupid’s matching algorithm with a critique your own relationship, well, let’s just say it would be you who is failing the logic test.

  10. 50

    Whoa, I found this video by googling “What’s wrong with OKCupid’s matching algorithm”, an article I wrote a few years ago that got on Hacker News. Seems like the video makes the exact same points that I made there, with no credit!
    There’s nothing wrong with the log scale (1,5,25,…) as this gives the user order-of-magnitude control rather than overly fine-grained control.

  11. 51

    I can’t believe no one pointed out the most obvious flaw in EMK’s argument on point 2. The difference between assigning 1 for a little important and 0 for irrelevant is not small, it is infinite. Multiplying by 0 means it will not be counted at all, which is what you want.  

    As others pointed out: no, people do not know what they want, but you can use the questions to filter (for drug use and smoking for example). This does not have to become an exercise in narcissism at all, since you can choose the answer they should provide.  So if you want to date someone who is opposite to you, you can choose to do so.

    So although EMK raises some valid points here and there, I really think he should rewrite this article, as he clearly did not understand the okc video that well. No offense. 😉

  12. 52

    The problem with okcupid and their “algorithm” is that the okcupid system does not know attraction nor does it calculate it into the search. Like for example I saw a person on there who was very attractive like one of the most attractive people I have ever seen in my life but our match %? A whooping 19%. Our enemy %? Close to 80%. So here I am with someone very attractive who I am basically non compatible with well not very much anyways. Then there was the time I saw someone as close to 95% match to me but when I reviewed their pictures I found I was not attracted to them at all. What I want is for okcupid to give you someone with a high match % who you can feel attracted to when you see their pictures. But I understand that would be virtually impossible to do.  I also wish they had a way to filter out results so that they only show people that have more than a certain match % so like say I could select it so that anyone under 75% doesn’t show up in my searches or something. I also put on my profile I am seeking someone 20- 38 (I’m 26) but I keep getting people in their teens and people above 38 mixed with my searches you would think their system would know based on the info you selected on your profile. But it is what it is I guess.

Leave a Reply

Your email address will not be published. Required fields are marked *