Monday, August 24, 2009

Unlucky number 37.

I discussed this briefly with one or two people, and I don't think it came out quite right.. here's the explanation on the odds and statistics of marrying the right (or wrong) person (from a great article in the New Scientist)


If you have trouble knowing when to quit, try getting your head around "diminishing returns" - the optimal stopping tool. The best way to demonstrate diminishing returns is the so-called marriage problem. Suppose you are told you must marry, and that you must choose your spouse out of 100 applicants. You may interview each applicant once. After each interview you must decide whether to marry that person. If you decline, you lose the opportunity forever. If you work your way through 99 applicants without choosing one, you must marry the 100th. You may think you have 1 in 100 chance of marrying your ideal partner, but the truth is that you can do a lot better than that.

If you interview half the potential partners then stop at the next best one - that is, the first one better than the best person you've already interviewed - you will marry the very best candidate about 25 per cent of the time. Once again, probability explains why. A quarter of the time, the second best partner will be in the first 50 people and the very best in the second. So 25 per cent of the time, the rule "stop at the next best one" will see you marrying the best candidate. Much of the rest of the time, you will end up marrying the 100th person, who has a 1 in 100 chance of being the worst, but hey, this is probability, not certainty.

You can do even better than 25 per cent, however. John Gilbert and Frederick Mosteller of Harvard University proved that you could raise your odds to 37 per cent by interviewing 37 people then stopping at the next best. The number 37 comes from dividing 100 by e, the base of the natural logarithms, which is roughly equal to 2.72. Gilbert and Mosteller's law works no matter how many candidates there are - you simply divide the number of options by e.


No comments:

Post a Comment