Beer consumption is generally treated as a matter of intuition. This seems surprising. The activity is common, repeatable, measurable, and exhibits an obvious diminishing-marginal-returns curve. It should therefore admit of optimization.
\nLet U(n) denote the utility of consuming n beers over the course of an evening. I model this as:
\nU(n) = S(n) + E(n) + C(n) − H(n) − R(n)\n\nS = sociability benefit\nE = euphoria\nC = conversational fluidity\nH = hangover cost\nR = probability-weighted regrettable behavior\nA major complication is that “one beer” is not a stable unit. A 12 oz lager and a 16 oz 9% double IPA plainly cannot both count as n = 1. I therefore normalize all intake into Standardized Beer Equivalents (SBEs).
\nEarly result:\nΔU1 > 0\nΔU2 > 0\nΔU3 > 0\nΔU4 ≈ 0\nΔU7 << 0\n(“Texting ex” becomes endogenous.)\nThe folk heuristic “a couple” turns out to be surprisingly robust, but too vague for high-stakes use. My current best estimate is that the optimal intake is n* ≈ 2.7 beers. Since fractional beers are physically realizable, I see no reason to round.
\nSome readers will object that this model does not sufficiently weight tail risks. I disagree. After a decade of field research, several lost jackets, one misdemeanor, and a continuing ban from McGlinchy’s, I have unusually high confidence in my calibration.
\nUpdate: multiple commenters persuaded me that drinks/hour may matter more than drinks/night. I therefore tentatively endorse the policy: “Know your limit, then exceed it by 0.7.”
\n","excerpt":"Beer consumption is generally treated as a matter of intuition. This seems surprising. The activity is common, repeatable, measurable, and exhibits an obvious diminishing-marginal-returns curve.","tags":["Decision Theory","Self-Experimentation","Alcohol","Applied Hedonics"],"score":398,"displayOrder":390,"thumbnail":"/assets/articles/optimal-number-of-beers-thumb.webp","image":"/assets/articles/optimal-number-of-beers.webp","commentCount":16,"date":"2026-05-28T00:00:00.000Z","updatedAt":"2022-12-03T00:00:00.000Z","revision":13,"status":"published"},"comments":[{"id":"seed-optimal-number-of-beers-8","author":"rate_not_total","text":"Drinks per hour matters more than drinks per night. Your body clears alcohol at a roughly fixed rate, so what you feel depends on intake against clearance. 2.7 SBEs over four hours and 2.7 SBEs in forty minutes are two completely different evenings, and the model can't tell them apart.","parentId":null,"score":112,"date":"2026-05-28T00:40:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-0","author":"BayesBender","text":"The post would be stronger if it controlled for the fact that Beer 4 is usually chosen by a meaningfully dumber version of the person who decided to have Beer 4.","parentId":null,"score":96,"date":"2026-05-28T01:00:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-1","author":"kalman_lager","text":"Related, have you tried modeling it recursively? The agent selecting Beer 5 is not the same agent who endorsed the Beer 5 policy ex ante.","parentId":null,"score":81,"date":"2026-05-28T01:20:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-9","author":"first_order_kinetics","text":"Seconding this. Every term in U(n) is really a function of the blood alcohol curve over the night, not the count. If you want one number per evening, use the area under that curve, or the peak if you mostly care about R.","parentId":"seed-optimal-number-of-beers-8","score":64,"date":"2026-05-28T01:30:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-10","author":"drinkmaxxer42","text":"Good point. I've added a correction term for within-night cognitive degradation, though unfortunately the correction was derived at a bar.","parentId":"seed-optimal-number-of-beers-1","score":43,"date":"2026-05-28T02:00:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-11","author":"standard_drink_pedant","text":"Standardized Beer Equivalents already exist, they're called standard drinks, about 14 grams of ethanol in the US. A 16 oz 9% double IPA is roughly 2.4 of them by itself, so a \"two-beer\" night with two of those is nearly five.","parentId":null,"score":88,"date":"2026-05-28T02:00:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-2","author":"drinkmaxxer42","text":"This is persuasive. I'm going to reformulate around rate and update the post.","parentId":"seed-optimal-number-of-beers-9","score":47,"date":"2026-05-28T02:20:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-3","author":"drinkmaxxer42","text":"I'm aware. SBEs are anchored to the experience of one ordinary lager, which is the unit people actually reason in. Grams of ethanol is not how anyone thinks at a bar.","parentId":"seed-optimal-number-of-beers-11","score":29,"date":"2026-05-28T03:00:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-7","author":"quesadilla_maker","text":"R is underspecified. The last time I went past four I woke up next to a quesadilla maker I don't remember buying. It was still plugged in. I don't know what term that goes in.","parentId":null,"score":77,"date":"2026-05-28T03:00:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-4","author":"direct_action","text":"Why not stop at the point where you're having the most fun?","parentId":null,"score":61,"date":"2026-05-28T04:00:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-5","author":"drinkmaxxer42","text":"The point of maximum fun is only identifiable in retrospect. You need the model to locate it in advance.","parentId":"seed-optimal-number-of-beers-4","score":38,"date":"2026-05-28T05:00:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-12","author":"behind_the_bar","text":"I tend bar at a place a lot like McGlinchy's. Your 2.7 is a reasonable number, but nobody orders 0.7 of a beer. In practice a fractional beer gets rounded up by whoever is pouring it.","parentId":null,"score":49,"date":"2026-05-28T06:00:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-13","author":"limit_creep","text":"\"Know your limit, then exceed it by 0.7\" is a policy that moves your limit. Run it for a few months and your limit is wherever you were last Friday plus 0.7.","parentId":null,"score":36,"date":"2026-05-28T09:00:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-6","author":"drinkmaxxer42","text":"That's a fair concern about drift. I'll recalibrate the limit quarterly against a fixed reference night.","parentId":"seed-optimal-number-of-beers-13","score":22,"date":"2026-05-28T10:00:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-14","author":"tuesday_vs_friday","text":"H also isn't a function of n alone. Same beers on a Tuesday before a 7 a.m. standup and on a Friday with nothing until noon are very different hangovers. The model needs a sleep term.","parentId":null,"score":31,"date":"2026-05-28T12:00:00.000Z","own":false,"moderatable":false,"seed":true},{"id":"seed-optimal-number-of-beers-15","author":"gigachad_utilitarian","text":"just drink 12 and find out","parentId":null,"score":-9,"date":"2026-05-29T00:00:00.000Z","own":false,"moderatable":false,"seed":true}]}