This week, I was introduced to Hopleaf, a craft beer bar in Chicago with 60 rotating taps. One of my favorite bars in Tucson, 1702, was similar. But every time I visited, I would face a dilemma: I usually only visit twice a month, and drink 3±1 beers each time. As a reformed economist, I insist on optimizing every decision I make, and with each beer I order, I face a trade-off between tasting a new beer of uncertain quality or choosing my favorite so far. If I want to maximize expected utility for each month’s visits, what is my best strategy? How many beers should I spend exploring for better ones, and how many should I spend enjoying my favorite?

Richard Feynman posed this problem in the context of finding the best dish at a new restaurant. (But beers are way more interesting). The solution makes sense (and come on, it’s Richard Feynman), but I wanted to check his work with a simulation.

You can find the results (and the rest of this post) in this iPython notebook. There’s much more there, including an interesting result!