OkCupid meets speed-dating

[gusl]

OkCupid is really good at measuring similarity and predicting friendship, but bad at predicting chemistry. I've met several good friends through it.

Speed-dating is high-throughput but low-yield (nearly random sample of Vancouver people). Experiments suggest that I'm compatible with <5% of Vancouver women 25-35, even for just Platonic friendships. (Take away the Platonic, and this probably goes under 1%).

Next Saturday I'm meeting with a number of men and women for dinner, matched through OkCupid.

Does anyone know if their compatibility score is a *metric*, i.e. do friend- and match-percentages obey the triangle inequality?

Comments

[widdertwin.livejournal.com]

I've always found those sites kind of silly and inherently inaccurate. I've taken several of such tests for romantic and platonic compatibility and compared my results against those of close friends as well as my boyfriend, and if they're anything to go by, I'd have little to nothing in common with any of them. With the exception of one of my very close friends, I'm pretty much the polar opposite of all of them. So much for the interwebs

On a totally unrelated note, are you going to be around in the afternoon this coming Saturday? I'm going to be on your end of town to pick up a vuvuzela from the Vuvuzela Lady, so if you're down for a coffee or something that'd be cool.

[gustavolacerda.livejournal.com]

OkCupid is different from its predecessors in the same way as Google, i.e. it actually works! My experience with other matchmaking -type sites had been mostly bad, so I found OkCupid quite remarkable. I think their secret is that they employ people who know something about math and stats, and are actually serious about their models.

Yeah, totally, give me a call when you're around.

[widdertwin.livejournal.com]

I suppose the experience you have depends on the purposes you're using it for. I've never taken any of it seriously in terms of finding potential friends or romantic partners, so perhaps that's why my results are the way they are; cynical results for cynical purposes.

Anyway, sounds good. I'll let you know when I'm in the area. You'll hear my obnoxious, interminable buzzing from a mile away.

[tedesson.livejournal.com]

You're such a geek!

If the chance of finding someone in the random population is <1%, spend more time looking in enriched populations.

Also, aim to acquire some skill/talent which distinguishes you from other geek men (your real competition). Gourmet cooking, for example, or a large social network that you inspire to do things together. The skill should be hedonic, not 'practical'. This is because geek men often have a hard time expressing affection in the real world, and a real physical skill is something to point to and say "Cool", and "Wow", and remember.

[candid.livejournal.com]

I'm guessing that it violates (at a minimum) the "d(x,y)=0 only if x=y" part of the definition.

[en-ki.livejournal.com]

I would be astonished if it didn't look like correlation, which (as we've discussed) is an inner product, which induces a metric but is not itself a metric.

[gwillen.livejournal.com]

I don't know if you've actually tried OKCupid, but to provide some anecdata for you: while I've never used it to find a partner, those I have retroactively looked up on it have scored in the high 90s on "match rating" -- high enough that, had I been looking exclusively on OKCupid, it's entirely plausible I would have found them there.

It was a little disconcerting, honestly.

[widdertwin.livejournal.com]

Am I the only one who finds matchmaking sites perverse?

[http://users.livejournal.com/_tove/]

Second. Everyone I know and like who has an OKCupid profile matches me fairly highly (especially past/current significant others), and the people I've met from OKC who I've most liked have been high match ratings.

The one person I met that I actively disliked had a lower rating (in the 60-70% range) and I only met because they seemed more my type of person in their email correspondence than their profile made them out to be. So OKC's guess was actually much more accurate than my own.

[gustavolacerda.livejournal.com]

ah... but if we reinterpret x=y to mean "x and y have identical answers", then it is a metric again.

[gustavolacerda.livejournal.com]

<< spend more time looking in enriched populations. >>

yup, that's the whole idea of "speed-dating" using our OkCupid networks.


<< Also, aim to acquire some skill/talent which distinguishes you from other geek men (your real competition) >>

I don't have a competitive mindset in this context / can't be bothered.

[gustavolacerda.livejournal.com]

That is also my feeling.

But could their matching technology be that simple? Just a correlation between profiles, reweighted by assigned importance? That would make it asymmetrical already, unless they have a way of symmetrizing it.

[http://users.livejournal.com/_tove/]

That probably depends on what you are trying to get out of them. I agree that they kind of support the paradigm of coupledom being superior to singledom, but a lot of people do genuinely use OKC (at least, and I can't speak for other sites) to make friends.

As someone who (at best) is self-employed, doesn't drink, and dislikes social occasions that involve lots of people, I don't have a whole lot of places to meet new people. (Work, bars, and parties are all out, and they're sort of the big three for most people. And most of my hobbies are pretty solitary.) Couple that with shyness, especially around girls, and I have a lot of difficulty making new friends. OKC does actually do a good job at predicting who I'll like and has a hell of a lot better quality control than, say, Craigslist. The fact that there's a certain amount of built-in reassurance of compatibility (and advance knowledge of my/their romantic situation; I don't have to be like, "this is strictly platonic" every time) helps make it easier to approach strangers. So, it doesn't seem perverse to me.

[gustavolacerda.livejournal.com]

having said that, I am trying to acquire some cooking skills.

[http://users.livejournal.com/_tove/]

So, they certainly claim to "record everything you do on this site." I have always assumed they also take into account things like how often you tend to view the full profile of (or message) users similar to the other profile -- the "customers like you also looked at" thing that Amazon does. I know they measure how often people with low match ratings to you message you (the "beauty" rating).

[darius.livejournal.com]

I get the impression people generally score below 100% match with themselves -- you'd have to be looking for the same answers as you give. So, not a metric.

[(Anonymous)]

Well, since the only people from the OKcupid part of the invite who showed up were guys looking for girls, I think we may still have failed you on chemistry. Good thing I got some 'real-world' friends to come. There was good food & good conversation, anyway.