BAC Calculator

Why this page refuses to give you one number

Every blood alcohol calculator on the internet is doing the same arithmetic. Most of them finish by printing something like 0.062% — three characters of decimal precision on a quantity that nobody in the room can measure. That formatting decision is not cosmetic. A reader who sees 0.062% next to a legal threshold of 0.08% draws a conclusion, and the conclusion is wrong, because the true value for that person at that moment could plausibly be anywhere from 0.045% to 0.090%.

The uncertainty is not a flaw in the implementation. It is in the equation itself. The Widmark model needs a factor r, the fraction of body mass that alcohol distributes into, and r is a property of body composition that weight and sex only roughly predict. Published values run from about 0.55 to 0.82 in men and 0.44 to 0.66 in women. Because r sits in the denominator, that spread alone moves the answer by nearly 50%.

Then there is elimination. The liver clears roughly 0.015 percentage points per hour, but the real interval across healthy adults is about 0.010 to 0.020 — a factor of two. That uncertainty compounds: it barely matters ten minutes after a drink and matters enormously six hours later, which is exactly why the band on the chart gets wider as the night goes on rather than staying a fixed width.

So this tool reports the band. Both edges are computed from the same equation with the plausible extremes of both parameters. The width of that band is the most useful thing on the page, because it is usually wider than the distance between your estimate and any legal threshold — which is the whole reason an estimate cannot answer the question people come here to ask.

The Widmark equation, and the part that gets hardcoded

Erik Widmark published the model in the 1930s and it is still the backbone of forensic alcohol calculation:

BAC = A / (r × m) − β × t

• A — grams of pure ethanol consumed.
• m — body mass.
• r — the Widmark factor, the share of body mass that behaves as a solvent for alcohol. Alcohol dissolves in water, not fat, so r tracks total body water.
• β — the elimination rate, in BAC points per hour.
• t — time since drinking started.

This page computes it in grams per kilogram of blood, the unit Widmark actually worked in, and then multiplies by 1.055 — the density of whole blood — to reach the g/100 mL figure that American law is written in. That conversion is skipped often enough to be worth naming: leaving it out makes every result about five percent low, and low is the direction that gets someone into a car.

Why women reach higher concentrations from the same drink at the same weight comes straight out of r. Women carry proportionally more adipose tissue and less body water, so the same dose of alcohol is diluted into a smaller volume. It is not a difference in tolerance or in how fast the liver works; it is arithmetic about where the alcohol goes.

Alcohol leaves at a fixed rate, not a fixed fraction

This is the single most common modelling error in online calculators, and it is worth being precise about. Most drugs are cleared by first-order kinetics: a constant fraction of what is present disappears per unit time, which produces a half-life and an exponential decay curve. Alcohol does not behave that way.

Alcohol dehydrogenase saturates at concentrations far below any level that would concern a driver. Once saturated, the liver removes a roughly constant amount per hour regardless of how much is in circulation. That is zero-order kinetics, and the descending part of the curve is a straight line, not a decaying exponential.

The practical difference is large and it points the dangerous way. An exponential model with a plausible half-life drops quickly at first and then flattens into a long tail that looks like "basically nothing left" hours before that is true. The straight line keeps descending at the same rate right up until it hits zero. Someone using the exponential version to decide when they will be clear will decide too early, every time, and by more the more they drank.

Zero-order kinetics also explains something people find counter-intuitive: doubling the amount you drank roughly doubles the time to clear it. There is no diminishing tail. Six drinks do not take a bit longer than three; they take about twice as long.

The peak arrives 30 to 90 minutes late

A calculator that treats the alcohol as instantly present in the blood and then subtracts elimination from the first sip has two errors that partly cancel and partly do not. It overstates the amount in circulation immediately after a drink, and it understates the peak, because it has already subtracted elimination time that had not yet happened for alcohol that had not yet been absorbed.

Real absorption goes through the stomach and small intestine and takes time. On an empty stomach the peak typically lands 30 to 45 minutes after finishing a drink. After a full meal, the stomach empties more slowly and the peak can be pushed out past 90 minutes, and arrives lower and flatter.

Two consequences follow, and both matter more than the exact number. The first is that checking an estimate right after your last drink is checking a value that has not peaked yet — you may be reading a rising curve and interpreting it as a final answer. The second is that food does not remove alcohol. It changes the shape of the curve, not the area under it. The same grams still have to be cleared at the same rate, and the total time to zero barely moves.

This tool models absorption as a first-order process whose rate depends on what is in your stomach, so the curve rises rather than jumping, and the elimination line only consumes alcohol that has actually made it into the blood.

Grams of alcohol, never "number of drinks"

A "standard drink" is a public health unit, not a physical quantity, and it is not the same size in different countries:

So this calculator never asks how many drinks you had. It asks for volume and ABV, and converts: grams = volume (mL) × ABV/100 × 0.789, where 0.789 g/mL is the density of ethanol. A 355 mL can at 5% is 14.0 g. A 473 mL craft IPA at 6.5% is 24.2 g — one glass, but 1.7 American standard drinks. A generous 250 mL pour of 13.5% wine is 26.6 g, nearly two.

This is where most self-assessment goes wrong, and it goes wrong in the direction of underestimating. People count vessels. The vessel has been getting bigger and the beer has been getting stronger for twenty years.

What 0.08% means in United States law

The 0.08% figure is a per se threshold. That is a specific legal concept: at or above it, the offence is established by the reading itself, with no need to prove that driving was actually impaired. It is a rule about what the prosecution has to demonstrate, not a finding about where danger begins.

Three things about that table are routinely missed. First, being under a threshold does not close the matter: every state can charge impaired driving on observed behaviour — weaving, field sobriety performance, an officer's testimony — at any concentration at all. Second, zero tolerance for drivers under 21 means what it says; in most states any detectable reading is the offence.

Third, and least understood: implied consent. By holding a licence you have already consented to chemical testing when lawfully arrested. Refusing is not a free move. It triggers an automatic administrative licence suspension in every state, frequently longer than the suspension for the DUI itself, it is admissible against you in many jurisdictions, and in some states it is a separate criminal offence. A refusal also does not prevent a warrant for a blood draw.

Nothing on the internet's list sobers anyone up

Coffee, a cold shower, a walk in fresh air, exercise, energy drinks, vomiting, greasy food afterwards, sleeping for an hour. None of them change the elimination rate, because none of them change how fast liver enzymes work. The rate is what it is.

Caffeine is the actively dangerous one. It raises subjective alertness without touching objective impairment, which produces a person who is measurably worse at driving and measurably more confident about it. The research on caffeine plus alcohol is consistent on this: perceived impairment falls, actual impairment does not.

Sleep is often assumed to be a special case. It is not, in the sense that matters — the liver works at the same rate whether you are awake or not, so five hours of sleep clears the same alcohol as five hours awake. It does not clear it faster. And on top of whatever alcohol remains, you now have sleep deprivation, itself a well-documented driving impairment.

What the estimate cannot see, by name

The equation takes weight, sex, grams and time. Here is the list of things it does not take, each of which can move the real value more than the difference between the two edges of the range:

• Body composition. Two people at 80 kg with 15% and 35% body fat have substantially different total body water and therefore different r.
• Drinking pace. Four drinks over four hours and four drinks in forty minutes produce very different peaks from identical totals.
• Stomach contents and timing. Not just whether you ate, but what and when. Fat and protein slow gastric emptying more than carbohydrate.
• Medication. Benzodiazepines, opioids, sedating antihistamines, muscle relaxants, sleep aids and many antidepressants add impairment that no BAC figure reflects. Some also compete for the enzymes that metabolise alcohol.
• Genetics. ALDH2 and ADH1B variants — common across East Asian populations — change both the speed of clearance and the intensity of effect, sometimes dramatically.
• Liver condition and health. Disease, illness, dehydration and fatigue all shift elimination.
• Tolerance. A regular heavy drinker may feel less at the same concentration. Feeling less is not being less impaired; reaction time and divided attention degrade on schedule regardless.
• What was actually in the glass. A home or bar pour is not a measured pour, and carbonation speeds absorption.

Everything on this page runs in your browser as arithmetic. Nothing is uploaded and there is no account. Use it to understand why the number is unknowable — that is what it is good for. Then take the ride.