Wednesday, April 22, 2009

the socks problem

I have the habit of leaving my clean clothes on the dryer until they run out or until I need to dry some more, case in which I transfer everything to a large basket. I rarely spend the time to put them away. When I need some socks I go to the dryer and try to find a matching pair. This process is often more difficult than I'd expect and it has bothered me for years.

So today I decided to compute the probabilities involved and the expected number of draws until I find a matching pair.

I draw one sock at a time uniformly at random until I have a matching pair. If I have 10 pairs of socks (all of them different), the probabilities of drawing a sock exactly at the k-th draw are as follows:

p(1 draw) = 0 (can't have a pair with only one sock)
p(2 draws) = (1 / 19) (I have 1 sock in my hands and I can draw any of the other 19 socks)
p(3 draws) = (1 - p(2)) * (2 / 18) (To get here, I can't have drawn a pair in the previous two draws; and now I have 2 socks and I can draw any of the other 18 socks)
...
p(10 draws) = (1 - p(2) - p(3) - ... - p(9)) * (9 / 11)
p(11 draws) = (1 - p(2) - p(3) - ... - p(9) - p(10)) * (10 / 10) (To get here, I can't have drawn a pair in the previous ten draws; I've also got all possible socks so anything that I draw at this point will match one of them, so the probability of finding a match here is 1)

The expected number of draws to find a matching pair is:
1 * p(1) + 2 * p(2) + 3 * p(3) + ... + 10 * p(10) + 11 * p(11)

That gives around 5.7 draws.

Ok, I don't feel as bad now. I still feel the universe is playing tricks on me, but at least they're subtle enough to keep my paranoia at bay.

Bonus: a graph showing the probabilities of finding a match on the k-th try.

10 comments:

hermann said...

I have the perfect solution that avoid all this problem

Have the same model for all your socks. But if you want different kinds, use different colors for each model, this way your color perception will considerably improve your selection.

Renata said...

I was just about to comment the same thing Hermann did. BUT you are not taking into account that you can LOOK inside the dryer when drawing socks! Or do you really "draw one sock at a time uniformly at random until you have a matching pair"??

BTW, you are SUCH a nerd :D

On a random note, I am going to start working as a Business Analyst for Dell, starting next month. Yay! :D

Thiago said...

All models are simplifications; you can't expect mine to be any different.

HOWEVER, given that I wash/dry my clothes in batch, usually there's a ton of other stuff in the dryer. Hence it's pretty usual for me to only find one sock at a time.

Renata said...

Sure. Estava apenas implicando. Tu estás muito nerd mesmo, só comentaste a parte das meias e nem deu bola para o meu comentário da mudança de emprego :P

JJM said...

The socks problem has a generalization for creatures with more than two feet. Thiago's approach does not work, but the problem is now completely solved, after much very nerdy work, and it's mathematically very interesting. A write-up is available from me.

Thiago said...

JJM: who are you? :)

JJM said...

James Madden, LSU Dept. Math.

Thiago said...

Awesome. I didn't mean to be rude, I was just curious. How did a professional mathematician end up on this page? :)

Did you post your n-feet generalization anywhere? I'd love to take a look.

JJM said...

I treat my socks the same way Thiago does, and I got interested in the problem for the same reason he did. It proved to be more interesting that I originally expected. See: http://www.math.lsu.edu/~madden/socks12-12-09.pdf

I found Thiago's blog when I started wondering if other people besides me had the same interest, and I googled "socks, dryer, match, probability" or something like that.

Anonymous said...

Wow, thanks for sharing this. This is truly helpful. There is another really effective solution, as well.

http://www.ryanboatright.com/matchingsocks