How to use math to choose a spouse | Hacker News Reader

How to use math to choose a spouse

news.cnet.com

11 points_pius8 comments

So assuming half the population of the world are female, that would give me 3388940742. let's say 20% of them are single (no clue, arbitrary guess), that would give a pool of 677788148 potential partners. Now probably 50% of those are within my acceptable age range. So to get a 37% chance of finding the "best" one in the world, I'd have to date 125516323 women and then choose the next best one (which would take longer since I already have a pretty large pool of exes).

It should be noted that this method only maximizes your chance of getting the best one, it doesn't take into account that the second best is much better than the worst. 37% of the time you use this method, you're just stuck with the last one you meet.

If you want to get as good of a wife as you can, only pass up √n before choosing the next one that is better.

http://en.wikipedia.org/wiki/Secretary_problem

I met my wife in linear algebra.

>I do not believe they were married. Lol!

I don't understand what they mean by "the next best one" after having discarded X number of candidates. "Best" relative to what?

I believe the assumptions are that there is a reliable fitness function giving an overall score, but only one candidate can be evaluated at a time, and once you reject a candidate, you can never go back.

So "the next best one" means the first to reach a new high value after the initial run of X. (And note that the first X candidates have no chance -- no matter how high they score -- because you don't even start with any hint of the overall distribution.)

It should be pointed out that while the article sort of munged this point, it is a particular math problem in game theory that, like other game theory problems, has applications of interest well beyond what the problem is "about", whereas it's actually useless in the real world. (How often does Prisoner's Dilemma actually come up for you?) So, if it sounds like an unreasonable assumption that there is a universal fitness function, well, sure. It's a math problem, not a real description of real life. 100 potential partners isn't terribly realistic either. It's either much lower ("the set of people I have dated") or much higher ("the set of all people I have ever met that might be a suitable candidate", and remember, in the face of "love at first sight" that's a very large set for most people, well into the thousands).

I find it sort of annoying when people propagate that error, but meh, whatever.

In the partially ordered set of partners (ordered by <=), given some ordering {p_k}, choose p_N from {p_(X+1), ..., p_100} s.t. p_i <= p_N for all i in {1, ..., X}.

Reply on news.ycombinator.com