Hundredth Percentile Beer
In statistics, there’s a concept called “normal distribution” that describes the range a particular trait exhibits across a population. It could be human height or family income or, relevant to our interests, beer quality. The classic representation of this is the bell curve:
Assessing the quality of beer is an evergreen riddle, and one we beer fans spend a lot of our time thinking about. We tend to focus on the beers that live in the edges (or “tails”) of that beer curve—the very good or bad beers—but the middle should attract our attention, too. Have you ever noticed how beers rated on Untappd always seem to have somewhere between 3.5 and 4.2? That’s a demonstration of the bell curve in action. In a normal distribution, two-thirds of all beers would be effectively average (with half slightly above and half slightly below). Raters are generous, but if we viewed the scores of all the beers in the database, I suspect the distribution would fall into the classic bell—even if the median was skewed higher than the numeric midpoint. (That is, the median* Untappd beer has a rating of perhaps 3.8.)
Really good and really bad beers are not common. Assuming a normal distribution, more than two-thirds of beers are average. There’s another sizable chunk that are just above- or below-average, 13.5% in either direction, or 27% together.** We’ve now accounted for 95% of all the beers. What remains are the very good and very bad.
Out in that right-side tail of the bell curve are beers in the 100th percentile. They’re rare, but not shockingly so. Out of every hundred beers made, one is exceptionally good. There are 8,000 breweries in the US, and they make, what, fifty beers apiece on average? That means 4,000 are in that hundredth percentile. That’s a lot!
All of this came to me because in the next month or so, I’ll debut the 2020 edition of Portland’s Best Breweries. Each year it gets a little harder as the number of breweries in contention for those ten slots grows (off hand, I’d guess there are a couple dozen). The very challenge that makes this exercise so difficult is the reason it’s so useful to visitors: there are too many breweries, so how do you find the ones you’ll like amid the surfeit of choice?
One way I’ve started to think about it is looking to see how many of those really good beers a brewery makes. If a brewery makes a hundred beers a year, you’d expect them to have a single hundred-percenter. But what if they have two? Or five? That would be quite a brewery! We could also look at the other beers. Are they all over the map, or do they skew one way or another? How many one-percent beers are there?
That brewery with five hundredth-percentile beers would be especially good if many of the other beers were in that “above-average” range (call them eighty percenters), and few were in the low range. It would be a better brewery than one with five great beers, five bombs, and a lot of above- and below-average beers.
I also like the idea of hundredth-percentile beer because it’s not that rare. When we talk about exceptional beer, we’re usually thinking romantically of a “world-class” or category-defining example. That’s when you get further out in the tail, where you’re talking about the very best of the hundredth percenters. It’s highly imprecise because of the nature of subjectivity. There’s something nice about not having to go that far. It’s a decent framework for eyeballing quality. Asking if the pint in your hand is a one-in-a-hundred beer (or even top five in a hundred) is both a lot easier to answer and freed of the weight of considering whether it’s a <hushed-tone> legend. There are breweries out there that skew to the right of the bell curve, breweries that skew left, and breweries that manage to have an unusual numner of hundred-percenters. Those are the ones I care about.
Your thoughts?
* The middle point dividing a population—half are above, half are below—is known as the median. It’s different from the average, or mean. If there are a hundred people in a bar, half make forty thousand a year and half sixty, the mean and median is $50,000. If Bill Gates walks in, suddenly, on average, everyone’s a millionaire. But the median barely moves.
** We don’t need to get into standard deviations, but that second tranche of the population on either side of the middle two reflects traits that are statistically above- or below-average, but no more than that. They’re still unexceptional, if somewhat rarer.