Failed Fetish Theory
Among f18225's recent string of excellent entries was one about fetishes. As in: "Bob totally has yellow fever. He's got a huge asian fetish."

It seems like the popular take on fetishes is that they, and the people who have them, are weird in a negative kind of way. It's a bit of a stigma. I was going to write a post with a mathematical evaluation of fetish "strategy", but I had to make too many assumptions, the post got too long, and I got confused.

Instead, we only make the simple statement: for most fetishes, under reasonable assumptions, more than half of the population benefits.

First, let's define a fetish as "a heightened attraction to a (smallish) subset of the population". We limit these subsets to ones based on characteristics orthogonal to attractiveness. (e.g. there is no fetish "for hot girls".) In this scenario, for simplicity, we assume a heterosexual fetish. We partition the population into four groups: F = fetishists; (1-F) = all other people of that gender; OF = fetishized group; (1-OF) = all other people of that (opposite) gender. Here's how each group fares.

F: By definition, since this group has a heightened attraction to members of OF, people in (1-F) don't find OF quite as attractive. Thus members of F have less competition in succeeding with their most desired partners, members of OF.

(1-F): Members of F are more concerned with OF, leaving less competition for (1-OF), which is good for (1-F).

OF: Members are subject to heightened attention, and thus have a larger swath of partners to choose from.

(1-OF): Members are less desirable on average, so they lose out.

But (F + (1-F)) = all people of one gender, so (F + (1-F) + OF) > 50% of the population. So fetishes aren't so bad.

(* This evaluation isn't strictly true; you can come up with degenerate cases using bizarre values of F and OF. Have at it...)

Anyway, a couple of years ago, I did a little fetish-testing experiment. I sent out pictures of 10 (fully-clothed) women to a large number of my guy friends. These 10 women had a variety of hair colors, body types, ethnicities, and so on. I had the friends rank the women from 1 (best) to 10 (worst), and aggregated all of their votes to create an overall top 10 list. Then I compared each person's top 10 list with the aggregate, and computed the deviation. The larger the deviation, the greater that person's "fetish index" -- and the better off he was. For instance, if a guy had a high deviation, that meant that a woman he thought was hot wasn't generally recognized as hot (what you might call a fetish). So, he was in a great position: that woman wouldn't receive much attention (since most other guys didn't think she was hot), so he's got a better shot of getting her. Good for him!

My results? Improbably, my top 10 (which was the first I inputted into the system) was exactly the same as the aggregate top 10... which meant that every girl I thought was hot, every other guy did too. Great, say, if I wanted to become a model talent scout, but bad for real life.