What they don't say is that you're probably hotter than you look in photos.
A cognitive scientist at the University of Chicago explained why to me last winter. When we look at ourselves in the mirror, in any given session we tend to anchor on the time slice image that makes us look our best. That, we decide, is the "real" us.
Photographs, however, are a random sample of the various arrangements of light, angle, and facial expression that we can be found in. The median photograph of you is probably the best approximation of your physical attractiveness. But that wars with your self image, which is anchored on other, better combinations.
I don't know if I'd call photographs a "random sample". I don't like being photographed, so I probably look less happy in photographs than I do on average, and happiness makes you look good.
But what's more, evolutionarily we're not used to looking at people at a single moment in time. At the time when our cognitive machinery evolved, there were no still photographs. (It's the reverse of the semi-early Internet, where there were still photographs but no video.) Even a person who you meet for a second will get more information about how you look than they would from a photograph. And perhaps photos look repulsive because they're not moving, and for a long time anybody who wasn't moving was dead.
McArdle also writes:
Which, by the way, is probably the best gauge of how attractive you are; how attractive are the hottest people who want to go out with you? They're probably only slightly more attractive than you are.
This is tempting, but basically impossible. I agree that it's true, among heterosexuals, and in percentile terms -- a man who's more attractive than 60% of men is probably most likely to be with a woman who's more attractive than 60% of women. (I'm assuming that there are an equal number of heterosexual men and heterosexual women.) But we don't think in these terms. Do you have any idea what these hypothetical 60th-percentile people look like? Furthermore, in looking at who people actually are in relationships with a lot more things than just their appearance come into play.
The only thing to do is the following: say you're a man. Then think "what if I were gay? Then who is the most attractive man who would be willing to go out with me?" That might work... but it tells you how hot you are as a gay man, and gay men have different standards for attractiveness than straight women. (Similarly for women; straight men and lesbians have different standards.) For the sake of simplicity I have assumed in this paragraph that there are exactly two genders and everyone belongs to exactly one of them.
Perhaps you could also go to hotornot.com (wow, I haven't looked at that site in years!) -- where people post photographs of themselves and other people rate them on a scale of 1 to 10 -- but we don't know how the scores are distributed there, and more importantly these are photographs.
If you buy the "best time-slice" theory, this also explains why you might find your friends more attractive than the average person. (I know I do.) I spend more time with my friends than with a random person I see on the street (if I didn't, they wouldn't be my friends) and I probably anchor on to their best time-slices. The times when they looked good are stuck in my brain even when they don't actually look good. If you figure that how good one looks at any given moment is normally distributed around their "true" attractiveness, then the "perceived attractiveness" -- the maximum attractiveness I've ever seen them at -- should grow without bound. After I've seen the person n times, the most attractive they've ever been is around the 100n/(n+1) percentile. This sounds ridiculous; the quantitative truth of how we determine someone's attractiveness is probably more like averaging the top √k or so of the k timeslices we have of a person. You still get the "attractiveness growing without bound" problem, but it takes a lot longer to come. (It's probably a normal distribution, because attractiveness is a sum of many different things, so I wave my hands and say the magic words "Central Limit Theorem", but that doesn't matter.) Also, your friends probably share the same cultural affinities that you do, meaning that what they do to make themselves seem attractive is probably similar to what you do to make yourself seem attractive. To avoid cognitive dissonance, then, you have to assume they're hot in order to be hot yourself.