What Is *Your* Best Profile Picture

May 6th, 2010 by Christian Rudder

One of our most popular posts ever was about profile pictures and what works and what doesn’t. For example, we examined why two very similar-seeming pictures of two very similar-seeming people often get very different amounts of attention:

As a follow-up, we’ve made something that can generate a personalized analysis of anyone’s profile pictures, something that can make a “blog post” just for you. It’s designed to help you choose your best pictures to put on your OkCupid profile, Facebook page, or whatever—sort of a more scientific (and more useful) Hot-Or-Not. We’re calling it MyBestFace.

It Will Analyze Your Pictures

MyBestFace is very simple: you upload photos of yourself; you contribute a few votes to the system; and after enough other people have done likewise, you get a report about your photos. The report is detailed and very awesome. I’ll show you mine.

It starts, naturally, by telling me which of my pictures was the most popular:

It also tells me what kind of people liked my photos: overachievers, deviants, jocks, whatever:

And I get a nifty map showing where all the people who voted for me are. It looks like I have heretofore undiscovered fan in the Philippines. Sweet.

My favorite part of the report is the end, where it shows the other photos mine were matched up against and exactly which people chose my picture as the best.

If you’d like to examine a complete example report, one of our beta-testers has generously agreed to let you guys browse hers. You can do so here.

Try It Out

MyBestFace is something we thought would be fun. We’re not thinking it’s gonna be huge or anything, but we hope it might be useful to anyone looking to do for themselves what we do here on the blog: figure stuff out. Upload a couple pics and play around with it. And let us know if anything seems busted! Besides your face! Burn!!!

104 Responses to “What Is *Your* Best Profile Picture”

  1. 1l2Hawk says:

    Some of these folks needed a NEITHER button!

    Is anybody else getting the error “Unable to open the internet site MyBestFace…” and crashing out to an “IE cannot display this webpage” screen in about 8 out of 10 attempts to click on anything?

  2. joe says:

    You need a 3rd option called “PASS”. Without it you are unrealistically inflating the status of an image. I also think a PASS option would do well to help people know what pictures they should IMMEDIATELY REMOVE. This site is plagued by high angle shots and duck-face; and a pass option would supply a much needed panacea.

    Otherwise this is pretty amazing. Excellent work!

  3. poisonwindy94 says:

    It’s great and all,but why do I have to just pick men?? I like girls. I think you need to put up a preferancence box,cause I dont want to pick a guy.

  4. chumeia says:

    So far the results briefly gave me a few new hits but only as a result of my activity on the “My Best Face” application. Posting the picture that other voted for has not altered my real world hits in any way visible.

    I posted two photo groups. The first was strictly a bunch of Random shots. Not yet knowing how the voting worked. I also included a control shot that should have always been voted down, but was also a test to see what people genuinely thought of me. I thought the system simply ranked the photos. I didn’t know they were randomly matched with others 2 at a time. Anyway this first group, contained a photo of a Toad. I really expected all negative and even to have the whole lot thrown out. But no the results came in, and infact the photo had received mostly negatives but I could hardly believe that of the sample data I was shown, some one HAD voted yes for the Toad. My associate explained it as a Toad Prince complex. I laughed so hard I actually cried. So you have to think to yourself when presented with a photo of a Toad and Man, why would you vote for the Toad? LMAO

    The second batch I posted were the best shots I could come up with. These had been my profile shots over the years and some had been highly ranked on “Hot or Not.” When the results came back it was obvious that the results were in line with what OKC had been saying. Strict face shots were not getting attention. People favor a shot with more of you and Other stuff in the shot. But that being said its also been proven that shots of you holding a beer don’t get as much attention, so don’t go wondering why you in a bar with friends doesn’t make the money. I will not vote for any one holding a beer or a cigarette. Additionally when asking of I like a person, if they are listed as highly religious, or being a member of the NRA, I will also not vote for them.

    Now then some are reporting the system not working. I went in and tried to view the default results for the blog. The browser gave me a script warning that a script was causing the page to run slow. I let it continue. The page lagged so bad that I could not scroll. When I finally got to the bottom I discovered that the MAP results seemed to be the reason for the script error. So that means the nifty little map, is causing some trouble. If you get a script error just tell the browser to stop the script. That should help.

  5. One Five Two says:

    I have one suggestion: Give us 9 photos to choose from and let us multi-select the most appealing photos. This will more realistically conform to real-world selection behavior. Forcing us to choose from 2 photos doesn’t feel right, because I often wish for a “neither” button.

  6. ironsmith says:

    I agree with 1l2Hawk that some comparisons need a ‘neither’ button, but some also need a ‘both’ button. Many of the pictures were VERY difficult to choose between!

  7. J says:

    I’d like to have a “both” or a “neither” button. Sometimes, there’s two with incredibly attractive pictures and sometimes it’s the exact opposite.

    Plus, a “use my OKC profile pics” option would be great. Other than that, awesome new feature.

  8. Double date says:

    OMG which part of “MYbestface” don’t these women understand??


  9. kelly says:

    It might be good to take into consideration the caption that goes with a photo. often it adds important context or just more information that makes the photo more meaningful. maybe another study though.. thanks for the new toy!

  10. Kyle says:

    This is awesome. Features like this are why I love OkCupid.

  11. Dave says:

    Interesting tool – great fun, too! Have run it twice now with different photos.

    It would be nice to have an option to select the age range of the reviewers.
    At 57, I’m more interested in what women my own age group think, rather than the population as a whole.

    As a category, “Women 31+” is waaay too broad and represents only a handfull of reviewers. I’d rather wait a few days to get a representative sample of my age cohort.

    Here’s something I noticed that with my own and competing photos:
    Mid-range shots with some action or attractive backgrounds were often selected over simple head shots. At full-size, mid-range photos may be more interesting, but would not work as a main profile photo in thumbnail size.

  12. TxTruthOrDare says:

    Love this! Thanks for putting this together.

    I would love it if there was something similar for profile text, which would let you pick the best of two “My Self Summary” write-ups, without pictures.

    Keep up the great work!

  13. wilderbuff says:

    Some thoughts.. I really wanted a “neither/pass” button for about 10 pairs.

    My dating preference skews heavily based on body weight, which was almost always why I wanted a “pass” button.

    I hate seeing profiles without any clear body shot (all “myspace angles”), but I found myself voting for face-only shots *in preference to* full body shots that weren’t appealing, even when I had a good indication that both women had comparable body size. I wouldn’t actually date either, but because I’m forced to choose I come down on the side of optimism, *even though* I almost never go on dates with women that are lacking a clear body shot in their profile.

    I also found myself noticing trends in the photos I liked and didn’t like (voted for every girl in a green outdoors environment; voted against all women in bridesmaid dresses) and I wished there was a way I could provide instant feedback about the reasoning behind my vote preferences.

    Overall, I felt myself being mostly superficial in my date choices… some women had interesting pictures but were significantly less appealing than their counterparts. I suppose normalization can correct this, but I think that allowing some additional level of description about photos would make the process more relevant, and maybe help us remember that we’re sorting pictures for good dates, not just good lays.

  14. Keepiru_co says:

    I do agree there should be a neither button, but I don’t think there should be a both button, it is about selection after all. The times where you could pick either, but have to select one will tell you the most about the quality of your picture.

  15. grneyes42 says:

    Enjoyed taking part in My Best Face~~but I was put up against very young women and considering I’m 67 it was rough getting the results that were explained in your survey.
    Just a suggestion~~in the future would it be possible in some way get age groups grouped closer. The voters ages don’t have to change~~ that would get some interesting votes from all of us comparing people in their own generations.
    Thanks for reading this. I really enjoy ok cupid~~it’s the best of all of the dating sites.

  16. crispy_crab says:

    Photos with groups of people: WHO ARE YOU?
    Also, maybe pitting people against folks of the same age would be more fair.
    People who just post pictures of their torsos or other bits under the ‘best face': you guys are twats.

    Maybe if we see all of their pictures and rate them against each other as well that could work.

  17. jakerake says:

    There should be a way to go back and change what you’re tagged as. I chose a couple, but then later noticed one other one that I fall under pretty well. Can’t find a way to add that tag to my profile.

  18. pen>sword says:

    Love this feature. The sample is still somewhat small and would like to suggest that the more charity votes someone makes, the more votes that person should get in his or her own report.

    What would also be great is to have some way of customizing the querry. The normalization is great but since everything is being bloated up from smaller numbers it doesn’t seem to reflect where the actual votes went. Unless you’re results show above 65% acceptance over the board, you’ll only have maybe one or two pictures that will say “Love this photo” on average which can be misleading. The said photo may only be marginally better than the next best photo and the normalization will show that they are worlds apart.

    At this stage, perhaps a dot plot may be better suited?

  19. tdogg says:

    There is no neither button because it would ruin the experiment. You have to pick one of the two because it figures out which picture is best based on statistics of which you pick.

  20. Danny says:

    I think this experiment will yield extremely interesting yet obvious info. For example it has made my biases extremely obvious. Age, race, and overall attractiveness (particularly weight) are coming into play when rating. I try to rate the “best” picture. I take composition into account, creativity, clarity, etc. but yet again even a great picture of a extremely unattractive person is well…..unappealing.

  21. Mark says:

    What is the test re-test reliability of My Best Face?

  22. Dave says:

    A bug?
    Ever since I got the report, my votes have been slowly disappearing. I think I started with 177 total votes (for and against), and now there are 170.

    – If feasible, let our pics continue to be voted on after the report has been “finished” (they could show up less often than newer pics), at least for a little while. That way the results could get more reliable – with 16 categories, how statistically significant could the results for one category – especially if it’s one of the less common ones – be?
    – If feasible, use the personality characteristics you already know about us from the questions we’ve answered. That could be instead of, or in addition to, the categories we select.

    That’s all for now. In general, great job – it’s fun and useful!

  23. anon says:

    People who report images of Mohammed are wastes of sperm and egg and OKCupid staff SHOULD NOT listen to them

  24. napalm says:

    clicked a few links, got to:
    the strongest visually distinguishing factor i notice is that she seems to use the least makeup in her “best” photo.

  25. Dissinnick88 says:

    I was presented with 2 pictures

    1 A girl laying on a guys chest in bed
    1 Old lady with a cat

    I feel cheated, i have to choose the taken chick or the crazy cat lady?

    oh well, crazy cat ladies need love too hehe

  26. Brad says:

    Damn this thing makes me depressed as an individual who is not attractive.

    The photos that constantly come in top as photos that are my best face are the ones where you can’t even see my face that well. It’s very displeasing!

  27. Aaron says:

    I have to say, I like the idea, but not the rating system. You called it a “more scientific (and more useful) Hot-Or-Not.” I think Hot-Or-Not had it right.

    You can test this yourself. First off, it’s just a fact that you’ll get a lot of variance with your method because you have fewer levels. Here–just two. Hot-Or-Not–ten. I suppose that with the compare feature that the compared pictures average out to the average picture. But it seems to be adding another variable unnecessarily. This isn’t quite approval voting because there are paired off comparisons. This is like some distorted Condorcet meets Approval mutant. (I’m not saying this is useless, just less useful than it could be.)

    Hot-Or-Not will leave you with a cold rating, one uncomfortably near μ. OkCupid may put things gently and pretty it up for you. But the cold truth isn’t always pretty.

  28. Jeni G says:

    I think I tried this the day it came out. After finding out my results I immediately switched the picture I had been using for “my best face”. The results were astounding. I now get alot more hits and guys trying to IM me than before. I always knew that people liked that particular picture of me (it’s also in one of my Facebook albums and gets LOTS of comments) but I had no idea. What a great idea guys.

  29. Dude says:

    Re-test reliability is pretty good, but not perfect. I’m an insecure overanalytical narcissist, and I’ve done dozens of photos, including a few duplicates to check how reliable it is:

    One picture got 79 in one test and 81 in another, for instance. Other dupes:
    77 and 75
    73 and 72
    72 and 67
    72 and 63
    71 and 71
    65 and 65
    61 and 56

    I’d love to know if this correlates to chicks’ scores. Girls with similar scores are pretty nice lookin’.

  30. Dude says:

    And yeah, it’s made my biases very obvious. Dark hair > light hair. Piercings/tattoos >> party girls. I’m not as racist against blacks as I expected, but I am more racist against asians than I thought. Sorry, azn peepz. But a slender girl of any race always wins against a fattie.

    Dating would be so much easier if everyone just lost weight. Maybe OkCupid should open a weight-loss site on the side. I bet the majority of these chicks are actually pretty hot, underneath their double chins. If OkCupid can get their users to increase their attractiveness, it brings more people to their site, which makes more money for them, and more palatable dates for us, and the ex-fatties don’t die of heart disease. Everyone wins!

  31. Bill says:

    Awesome feature! I think the reason they are only comparing two is because there would be a small population of users of this feature and they want each photo to get enough results to move the needle (they may be applying some kind of hypothesis testing in the background who knows). Yea the multi-photo paradigm would be closer to the real world, but you need many more votes to make the results significant (unless you posted a dominating photo 😉 ).

    A cool variation for unattractive people might be a kind of A/B testing setup so they aren’t competing with a highly dominant population of photos, just with themselves.

  32. Chris says:

    I think that what you have done with this is awesome. I understand the complaints of people wanting more options and choices, but frankly, this way is probably best. Using simple yes/no data, with a lot of data points, makes the statistical analysis more rigorous, and probably more useful, I think. Without any real basis for this, I assume that subconscious preferences are more likely to be exposed here, with a better idea for what people really want. Keep up the great work!

  33. savannuhhh says:

    “I’d like to have a “both” or a “neither” button. Sometimes, there’s two with incredibly attractive pictures and sometimes it’s the exact opposite.”

    i cant stress how much you need this feature pretty pretty pretty pretty please.

  34. Raymond says:

    I’d like to post a suggestion for the test’s format:

    Instead of pitting one user’s -single- picture against another user’s -single- picture, I’m thinking the better test would be to let users -rank- another user’s -batch- of profile photos.

    The way the current format is, I think it’s kind of hard to gauge which photos people like more, because they’re pitted against other users’ pictures. And, well, the way they are pitted seem random, so what photos get favorable reviews does seem to have an element of randomness to it (although I do understand that some will get universally panned and some will receive generally good reviews).

    Also, it becomes more of “who do you like” rather than, “which of this person’s photo is the best”, because we only get to see one of the user’s photos.

  35. K88Special says:

    I have to say, I absolutely love OKTrends. I’ve read several of your reports, and being a statistician and psychology nut, reading up on what does and does not work, and the overload of data you provide (well, actually it’s quite digestible) is great. I don’t know how much it will actually help those in the online dating world, but all presentation of data is excellent. Keep up the good work!

  36. FluffyGreenMan says:

    So I’m thinking that privacy for the reports is a very good idea.

  37. Green says:

    Adding to the suggestions above, I think the phrase “choose one for a date” is not the best measure if we’re trying to pick the best *photo* of the two. Over time I found myself shifting my judging criteria away from whom I’d want to date (e.g. not interested in Asian girls across the board) to whom I thought had the better/more telling photo. For instance, I couldn’t in good conscience give the thumbs up to a girl’s photo posed a hundred feet away even if the competitor was totally distasteful to me.

    And another thing — Like should be paired with Like. Pitting a 20-something against a 50-something sitting with her children unfairly stacks the odds in one direction.

  38. Jeremy B says:

    I was looking at the new My Best Face software. I noticed some things about myself. For example, I usually do not pic photos wherein the woman is eating junk food. It seems childish. I have a “favorite pose”. It’s very much like the one described in this study. There are some things to add, though: the woman is touching her hair, the woman’s eyes are usually looking down, and the woman has puffy lips. I pick photos where the woman is closer to the camera, and when she is in an environment that I imagine I would enjoy. For example, I would pick a photo of a woman taken from a high angle with water in the background because I would think of swimming, over a picture of a woman in her room. Also, I never pick pictures of aggressive girls or girls who look like they want to club…. Now, since I described myself as being almost every check box….

    Well, perhaps it’s just me, right? I think that it would be worthwhile to check some additional criteria out and widdle things down. And do some case studies, especially (because NOTHING in psychology is better than a case study), to see when the exception beats the rule..

  39. ChameleonDave says:

    Oh my god, you people are clueless. You’re not actually choosing one for a date; you are just helping rank pictures. OF COURSE it often gives you ones that are hard to choose between: that’s the entire point.

    And in any case you can simply hit F5 to reload the page. That’s a way to pass.

    The ‘report’ thing is a bit vague though. It should be made clear that you ought to report every photo with more than one person in it.

    It was stupid that ‘use my OkCupid photos’ was not the initial default, with the option to use different pics added later. Instead, they did the opposite! It was absurd to have to re-upload files that you already had on your servers.

    It would be excellent if I could restrict this geographically. I’m not interested in what American women want, which is what I’ll get if I use this tool right now. My best picture is apparently one where I’m holding a koala, but of course koalas seem interesting and exotic to Yanks. I’m only interested in knowing what local Aussie chicks think of that photograph.

  40. ComposerNate says:

    For instances of “neither”, I clicked on 1 and 2 together, apparently voting for both.

  41. Polyergic says:

    I didn’t think of it until after fulfilling my quota, but at least for the “charity” votes you can effectively skip a vote by following the link to the mybestface homepage.

  42. soiducked says:

    What I don’t get is that instead of using the personality stats that OKC generates and/or your match% to determine what sort of people you’re attracting and whether they’re the kind you want, there are these self-determined categories. While I want to attract self-professed “geeks” with my photos, I care a lot more about attracting people who have a high match% with me! Can we please get this information collected and analyzed by the reports as well?

  43. feargus says:

    There’s something kinda unbalanced here. My GF tried this, and she got her response in less than 2 hours. I did it and 26 hours later, I still have no report finished. Perhaps, due to the demographic difference, androsexuals should be required to rate more people than gynosexuals to get their report.

  44. shy_at_first625 says:

    I would also like four buttons for each pairing (1,2,neither,both), but my big wish is for a better sampling. It looks like I’m only getting sample size 20 for each photo, and 30 is the minimum for statistics problems.
    I second the “show me people in my age range” and “sort by sexual preference” requests, because the people I’m most interested in meeting are sometimes not in the sample at all, and I really don’t care what non-straight women think of my photos.

  45. Confuuuused says:

    Am I being stupid, or have you removed the feature to use your OKCupid profile for this site?

    I am told I can use facebook, or upload photos… Last time I just clicked a button on my profile and it copied all my photos over to a report. It was awesome. Now it is a pain to use.

  46. Ana says:

    I think this is cool, but a suggestion I’d have for improving the system would be to adjust the settings for the physical “types” that people are biased towards picking… For me, a blonde dude has to be particularly attractive for me to pick him over a guy with darker hair. It’s a definite bias that I have… and I know from having grown up as a black latina in a mostly white, bougie, pedestrian, suburb on the west coast that I’m not everyone’s type. No matter how objectively cute a guy may think I am from my pictures, if he’s only into dating white women or asians or white latinas he more than likely will not pick me over them, no matter the quality of my pictures or how better looking or not than the other woman I may be. A preference is a preference… And the reverse is true as well, though, I think that more people will pick the pedestrian looks over the less pedestrian looks, since most people are pedestrian.
    So, I don’t know, if you could put a question for the people who are doing the voting, like “What physical type do you normally date?” or “What physical type do you prefer to date?” and then people could check off stuff like race, hair type, eye color, hair color, weight, etc., and then the people who are submitting the photos could also fill out a physical description of themselves using the same options to check off, you could possibly have some way to account for the bias people have for certain types. I think other people with less “mainstream” looks such as myself would agree.

  47. carpentron says:

    I’ve done this three times now with the same basic set of photos, and the results have been very drastically different each time. I mean, to the point that my best can become my worst and vice versa. It seems to me that it doesn’t have a large enough sample size to be a very accurate way of figuring out your best photo unless there is a clear and marked difference between the quality of your photos.

  48. Kristine Rose says:

    i think it would work better if you rated someone’s photos only against themselves. otherwise people are really voting for the attractiveness of the person in the photos and not the photos themselves.

  49. pmailkeey says:

    Noting the offering of quick photo selection using ‘1’ and ‘2’ – why not go the whole hog and have 1-5 for the left image and 6-0 for the right. Such that if you preferred the left image a lot, choose ‘1’ but if there wasn’t much in it, choose ‘5’ This would solve the neither/both request.

    I like the idea of connectingsingles’ option to vote on profile photos – to give an indication of what viewers like and dislike. HOWEVER that would normalise the photos and hence hide oddities of people.

    I also think that comparing photos would be a means for OKC to do more match evaluation. I can’t figure out how one could quantify photo attributes such that the data could be processed by a computer !

    Finally, I’ve not found a better ‘dating’ site than OKC. There are other good ones but I’m struggling to fault this one in any respect.


  50. Admiral Happy says:

    I would really like a refine button. where you can redo an existing report without losing the existing information. So you have a bigger data set, and a more accurate result. All for just another round of voting.