This is quite exciting: dating site OkCupid took half a million profile pictures, asked three million people to choose between two randomly chosen examples, and ran the stats on the resulting photos’ EXIF data.
No really. This is seriously interesting. Even taking into account my current state of mind. Because there’s lots of advice in photography, particularly when it comes to portraits, and it all seems valid and reasonable and important until you show a bunch of pictures to somebody and they immediately reach for the dodgy one. The one where you had no choice but to fire the flash in someone’s face. The one where you’re shooting a woman from below because some dude was in the way and that was your only option. The one where the subject is blurred, but it’s the only shot you have of that particular moment. And this obviously dubious photograph turns out to be someone’s favourite shot, because it captures something personal to them that you couldn’t possibly know. This seems to happen all the time. And you start to wonder how important all the advice really is.
I once saw a photographer’s email signature that said “photographers like composition, non-photographers like smiles”, and sometimes I wonder if that’s the best tip I’ve ever heard. What’s needed is some actual data. And OkCupid is the perfect place to look, as dating profiles are modern portraits’ raison d’être.
I’ll skip to the conclusion: it’s good news for photographers. Everything that should be true about portrait photography turns out to actually be true. OkCupid have proper analysis, but I’ll unashamedly yoink the headlines:
- SLRs take more attractive photos than point & shoots. Camera phones are a long way behind1.
- Direct flash sucks.
- Low apertures (which blur the background) are seriously effective.
- People look better during the golden hour – the soft, golden light just after sunrise and just before sunset.
None of which is a big surprise to portrait photographers, but this should be a big deal. The evidence up to now has always been anecdotal. Actual numbers = win. OkCupid’s methodology seems reasonable, with the standard internet-survey caveats, and the numbers are enormous. It’s nice when the humanities can play in grown-up world for a bit.
There’s lots more you could do with such a dataset. Another classic portrait tip is never, ever to use a wide-angle lens2 as it makes people’s noses look huge, amongst other distortions. This seems true, but I’d love to see the data. However, taking into account different sensor sizes would be a nightmare, and I can see why OkCupid didn’t go there. Similarly the actual distance from the lens might be interesting, as would white balance (does it make a difference if men have a reddish hue?) and colour analysis generally. I’d also like to see how this gels with the previous – much more surprising – analysis which showed the much derided ‘MySpace angle’ (taken by holding your camera above your head and looking sultry) is the most successful style of profile shot for women by a long way, even when you control for cleavage.
Also: I wonder if single photographers are ahead of the curve. Hmmm.
Anyway – it seems that, on average, the classic rules of portraiture are valid. This obviously doesn’t mean anything for the quirky outlier photos that strike an unknown chord, but it suggests confirmation bias is perhaps playing a large role in my memory – I don’t remember the times when people like the photo that follows all the rules, because obviously they would. Portrait photographers can get a ‘safe’ shot from the old formulas – then you can start to play 🙂
…oh, and apparently iPhone users have more sex3. The numbers don’t lie, guys.