We Tested Distilling Calculators Against NIST. Here Is What We Found

In August 2026 we audited all 40 calculators on this site line by line against primary sources. The uncomfortable result: four of our own formulas were wrong, and two of those errors were not originally ours. They were the hobby's errors, absorbed from tables and formulas that circulate from site to site without anyone checking the originals. That raised an obvious question: how do the other calculators the hobby relies on hold up against the same sources? This page reports what we could verify, shows the arithmetic, and gives you an anchor table to test any calculator yourself in two minutes.

Methodology: We Only Test What a Tool Publishes

An accuracy claim is only fair if it rests on what a site itself states. This study therefore uses two evidence classes: worked examples a calculator publishes on its own page, and methods a page describes in its own words. Several popular calculators publish neither; those are listed as not testable rather than guessed at. Where a tool gets the method right, we say so. Our own site appears in the failure column, because before the audit it carried two of the errors described below. All tools were tested as publicly served on 17 August 2026. If you run one of the tools mentioned and believe our reading is wrong or has since been fixed, use the feedback form and we will retest and correct this page.

Finding 1: Dilution Calculators That Ignore Contraction

Mixing ethanol and water shrinks the total volume: 50 mL of each yields about 96 mL, not 100. A dilution calculator that simply adds volumes therefore under-doses the water and overshoots the strength. The OIML R 22 alcoholometric tables, the same density data used for legal proofing, give the exact answer.

A live example. The dilution calculator at distilling-spirits.com publishes a worked case taking 3.71 litres of 58% spirit down to 43%, prescribing 1.29 litres of water for a final volume of exactly 5.00 litres, and the page states plainly that contraction was not considered. That honesty is to its credit, but the numbers that follow from the simplification are measurably off:

A third of a percent sounds small, but the error scales with batch size and shows up exactly where precision matters: bottling to a labelled strength, and any context where proof gets checked. A simple-mixing mode is fine for rough work, as long as the tool says which mode it is using and offers the corrected one when it counts.

Finding 2: Temperature Tables That Run Backwards

An alcoholmeter is calibrated at 20°C. Warm spirit is less dense, the instrument sinks deeper and reads high, so a warm reading must always be corrected downward. NBS Circular 19, Table 7, the primary source underneath every legitimate correction table, puts the slope at roughly 0.35 percentage points per degree in the mid range and about 0.29 near azeotrope strength.

The tables circulating on hobby sites frequently get this wrong in two ways at once: direction flipped, and magnitude about half the true value. We know exactly how that happens, because this site's own correction guide carried precisely such a table until August 2026. It was not derived; it was absorbed from versions already circulating, and it survived until checked against the Circular 19 anchors. The corrected table and its derivation now live in the temperature correction guide.

Test any correction tool with these anchors:

If a tool corrects an 84.0 reading at 30°C to anything above 84, the direction is wrong. If it lands near 82.5, the direction is right and the magnitude is half. The correct answer is 80.7.

Finding 3: Fermaid-O Doses Four to Five Times the Protocol

Nutrient calculators disagree with each other more than any other tool class we examined, and the disagreement has a single cause. Scott Labs' raw figure for Fermaid-O is about 40 ppm YAN per gram per litre, but organic nitrogen is utilised roughly four times more efficiently than inorganic YAN, which is the explicit basis of the TOSNA protocol published by its author at meadmaderight.com. A calculator running the raw mass balance without the efficiency factor produces four to five times the protocol dose.

The protocol answer lands inside the widely quoted 3 to 4.5 g per gallon range; the mass balance answer is a jar of wasted nutrient and a head start for spoilage organisms. This site's mead nutrient calculator ran the raw mass balance until the August 2026 audit. Both camps are live across popular calculators today, which is why two well-built tools can disagree by a factor of four on the same batch.

Finding 4: The Right Standard, the Wrong Table

This one is ours. In the thread that followed publication, a reader reported that the app they use answered the dilution case above with 1.31 litres rather than either the 1.29 of ideal mixing or our 1.33. Checking why led to a TTB Table 6 implementation, and comparing the two standards against each other led somewhere we did not expect: our own density table was the wrong one.

OIML R 22 publishes density at 20°C in two separate tables. Table IIIa gives density as a function of alcoholic strength by mass. Table IVa gives it as a function of strength by volume. ABV is strength by volume, so Table IVa is the one a dilution calculator needs. Ours followed the by-mass curve. The endpoints are shared, water and pure ethanol, which is why nothing looked obviously wrong, but the whole middle of the range was displaced by up to 19.6 kg/m³ at 70%.

The tell was sitting in the first line of this page. A correct table has to reproduce the contraction that opens Finding 1: 50 mL of ethanol plus 50 mL of water giving about 96 mL. TTB Table 6 says 96.2 mL. OIML Table IVa says 96.5 mL. Our table said 98.1 mL. Two independent primary sources agreed with our prose and disagreed with our arithmetic, and nobody had checked one against the other.

The effect on the published answers is small, because the error largely cancels when two strengths are differenced: 1.330 becomes 1.332 on the case in Finding 1. It grows with the size of the drop, reaching 0.13 L on 20 litres going from 80% to 46%. Where it genuinely misled was any statement about contraction itself, which is the claim this study opens with.

The fix is not a better transcription. OIML R 22 publishes the general formula its tables are generated from, the Wagenbreth and Blanke polynomial, so the table is now computed from that formula and checked against the published Table IVa values on every build. It reproduces them to 0.005 kg/m³. A number that is derived and self-checking cannot drift back to the wrong table silently, which a copied array evidently can.

One consequence worth recording: our answer and the TTB Table 6 answer now agree to 0.002 L on the case in Finding 1, unrounded. Before the correction they did not. Two independent standards converging is a stronger check than either one alone, and it is the check that was missing.

What Held Up, and What We Could Not Test

Fairness cuts both ways. The fortification calculator at philbillymoonshine.com describes its method as a traditional Pearson's square, which is the correct approach for that problem; tools that document their method this way earn trust before a single number is checked. Simple tool classes such as proof conversion and volume arithmetic were generally sound wherever the method was visible.

On the other side of visibility: the calculator pages at americanhomedistillers.com serve their tools without publishing formulas or worked examples, so their accuracy could not be assessed remotely and no claim is made about it here. Yield estimators described as splitting a wash into alcohol collected and water left behind, a framing used for example by the distillate-to-water tool at howtomoonshine.co, describe a perfect separation; real runs collect at 40 to 95 percent strength with losses in both directions, so treat that class of output as an upper bound rather than a prediction.

And the admission that anchors this page: before August 2026, this site failed two of the tests above, and until 30 August it was using the wrong one of two density tables printed in the same standard. The full audit, including what was wrong and which sources fixed it, is documented on the methodology page.

Test Any Calculator in Two Minutes

Paste these cases into any tool of the matching type. Each has one defensible answer from a primary source.

Dilution: 3.71 L at 58% down to 43% needs 1.33 L of water (OIML R 22). An answer of 1.29 is ideal mixing. An answer of 1.31 is TTB Table 6, which is not an error: it is the US standard, and the difference is mostly the two-decimal rounding the regulation prescribes in its own worked example. If you are proofing for TTB purposes, 1.31 is the number that counts.

Temperature correction: a reading of 84.0% at 30°C is truly 80.7% (NBS Circular 19, Table 7). Anything above 84 is corrected backwards.

Fermaid-O: 19 L needing 234 mg/L YAN takes about 28 g total on the TOSNA basis. An answer near 111 g is the raw mass balance.

If a tool you rely on fails one of these, tell its maker. These errors survive because nobody checks, and every corrected calculator makes the hobby's numbers a little more trustworthy.

Frequently Asked Questions