How Accurate Is the Farmers' Almanac? We Scored Five Winters

Sixteen of its 30 scoreable calls predicted above-normal snowfall, in a stretch where only about a quarter of station-seasons came in above normal. The failure in our window is not randomness. It is direction: the almanac kept calling for more snow than fell, during five winters that were mostly short of snow.

What would a rule that says below normal every year have scored?

Eighty-two percent, on exactly the same station-seasons. A rule with no skill, no method and no almanac behind it, one that simply says every region will finish below normal every winter, would have beaten the published forecast by more than 50 percentage points across this five-winter window.

That comparison is the fairest way we know to present the result, and it cuts both ways. It shows the forecast performing badly, and it also shows how easy this particular window was to score well on, which is a caution against reading our 29.5 percent as a permanent verdict on the publication.

Which forecasts could not be scored at all?

Nearly half of them. Of 53 region-seasons we examined, 23 named two directions at once, predicting above-normal snow in one part of a region and below-normal in another. That is 43 percent of the forecasts, and no outcome can contradict a statement shaped like that.

A representative example, from the Atlantic Corridor for winter 2021-22: above-normal snowfall in the north and below-normal in the south. The region runs from Richmond to Boston. Whatever the winter did, some part of that sentence was right, which is why we set those calls aside rather than scoring them as hits.

Was the Farmers' Almanac accurate last winter?

Partly. For winter 2025-26 the Old Farmer's Almanac called below-normal snowfall, and 66 percent of our stations finished below normal. At region level it went one for three: the Northeast call verified at all four of its cities, while the Upper Midwest managed two of four and Texas-Oklahoma missed.

Its strongest winter in our window was the one before. In 2024-25 it was right in three of five scoreable regions and at 18 of 23 individual cities, the only season of the five where it cleared half on either measure. The two that sank the average are 2022-23 and 2023-24, when it called above-normal snow across the East and the East recorded some of its lowest totals in decades.

The number of rows behind each season differs because the number of scoreable calls does. A winter in which the Almanac hedged most of its regions leaves fewer calls that can be marked right or wrong, which is why 2025-26 rests on three regions and 2023-24 on ten.

Why is the Almanac's own score so much higher than ours?

Mostly because of who picks the city. The almanac scores the direction of departure for one representative city in each region and selects that city itself. We scored every city it lists under each region, which for the Lower Lakes means eight cities from Chicago to Syracuse rather than one.

Those two approaches can produce very different numbers from the same forecast. A region-wide call for above-normal snow can be right in Syracuse and wrong in Chicago, Cleveland, Detroit, Milwaukee, Buffalo, Indianapolis and Rochester in the same winter. We are not accusing anyone of choosing conveniently, because we cannot see their selection. We are pointing out that the method leaves that door open, and that a reader has no way to audit it.

There is a second, simpler reason. Their 94.4 percent figure is precipitation, which counts rain. Ours is snowfall alone. A winter can deliver normal precipitation as rain and still finish far short on snow, and across these five winters that is close to what happened in much of the country.

How does the Almanac make its forecast?

By a formula it does not disclose. The Old Farmer's Almanac says its method descends from a secret formula devised by founder Robert B. Thomas in 1792, built on the idea that sunspots influence weather on Earth, and states directly that the exact formula is proprietary.

That is the central problem for anyone trying to evaluate it. A forecast method that cannot be inspected cannot be tested by anyone except through its outputs, which is what we have done here and what Walsh and Allen did in 1981. Both publications also decline to publish the underlying regional data that would let a reader reproduce their accuracy claims.

What does the Farmers' Almanac say for winter 2026-2027?

It expects near-normal temperatures across the Great Lakes, Ohio Valley and Midwest, and says plainly that this does not mean a mild winter. Its wording describes fast-moving clippers, sudden cold snaps, brief snowy periods, snow squalls and bursts of bitter cold, interrupted by milder breaks.

The forecast was published on 6 August 2026 and last updated on 3 September 2026. It gives no snowfall figure in inches for any American city, which is true of every seasonal almanac forecast we have read, and it is the reason these forecasts are difficult to score without the kind of station work described above.

What does the Old Farmer's Almanac say for winter 2026-2027?

It splits the country into 18 regions and gives them different answers, including two that disagree across a single state line. Its winter forecast page calls for near or below normal snowfall for most areas, with heavier amounts in a few named zones.

Northern Ohio sits in its Lower Lakes region, which it calls warmer than normal with near to below normal snowfall. Southern Ohio sits in its Ohio Valley region, which it calls colder than normal. We work through what that split means against 77 years of station data in our Ohio snow guide.

Both publications also share a driver this season, which is the strong El Nino NOAA has under an advisory. The Climate Prediction Center's ENSO discussion is updated monthly, and NOAA's own winter outlook does not appear until mid-October.

What is the Old Farmer's Almanac predicting for each region this winter?

Below is every regional snowfall and temperature call on its page as it stood on 4 September 2026, following an update dated 31 August. We reproduce them so you can check next spring whether they verified, which is the whole point of keeping this page rather than replacing it each year.

Five of the fifteen snowfall calls in that table are already hedged across two outcomes. The Northeast is given opposite directions in its north and south, and the Lower Lakes, Upper Midwest, Heartland and Pacific Southwest are each given a band spanning near and something else. That is the same pattern we measured across five past winters, visible in this season's forecast before the winter starts. Regions 17 and 18 are Alaska and Hawaii, which sit outside our table and outside the test.

Which Almanac region is your state in?

Most big snow states sit in more than one, which is why a single state answer often does not exist. The Almanac draws its region lines through states rather than around them, so Ohio, New York, Pennsylvania, Michigan, Texas, Wisconsin and Colorado each straddle a boundary with different calls on either side.

The table below places each state by the cities the Almanac itself lists under each region, which is the same city set our test used. It is not a reproduction of the Almanac's own map, and a state can hold ground in a region that none of its listed cities falls in.

Colorado is the sharpest case in that table. Denver is one of the cities the Almanac lists under the High Plains, which it has down for below-normal snowfall this winter, and Grand Junction is one it lists under the Intermountain region, which it has down for above-normal. Both cities are in Colorado, and the two calls point in opposite directions. Our Colorado snow guide works through what the station record says about that divide.

Every forecast we scored, and how it did

Here is the full working, so nobody has to take our word for the headline figure. Each row is one regional snowfall call from an archived forecast, the direction we coded it as, and how many of that region's own listed cities finished on the predicted side of normal.

Reading down the call column shows the problem more clearly than any percentage. All five hits came from the ten below-normal calls. The sixteen above-normal calls produced none, and neither did the four near-normal ones.

Should you plan a trip around an almanac forecast?

Not as a booking decision. A seasonal outlook that gives no figure in inches, covers a region several states wide, and in 43 percent of the cases we examined predicts both directions at once, cannot tell you whether your February drive through the mountains will be clear.

Where it does have value is as a conversation starter about a season's character, and both publications are honest that they forecast trends rather than events. The Old Farmer's Almanac states outright that nobody has gained sufficient insight to predict weather with total accuracy, which is a fairer description of the product than the 80 percent headline suggests.

What should you use instead?

Climatology for planning, and a real forecast for the week you travel. A station's 30-year normal tells you what a typical winter delivers where you are going, and the historical range tells you how far from typical it can land. Neither requires anyone to predict anything.

That is the approach behind our own winter guides, which pair NOAA station records with what forecasters are currently saying rather than issuing numbers of our own. Our USA winter snow predictions hub collects them, and the Michigan guide shows the method applied to a single state.

For the days that matter, use the National Weather Service forecast for the specific place, issued a week or less ahead. Seasonal outlooks and short-range forecasts are different products built by different methods, and a seasonal outlook is not a weak version of a forecast, it is a different thing entirely.

The limits of our own test

Five winters is a small sample and we would not defend the 29.5 percent as a permanent figure. The window was unusually low on snow, which punished a publication that leaned toward above-normal calls, and a different five winters would very likely produce a different number.

Four other limits belong on the record. We tested the Old Farmer's Almanac, because it is the one with archived forecasts we could retrieve, so nothing here measures the Farmers' Almanac of Maine directly. We tested snowfall only, so this is not a test of the 94.4 percent precipitation claim. Our coding of which calls counted as above, below or near normal is a judgement, which is why the verbatim wording of every call sits in our working notes. And the 1991-2020 normals may themselves lag a warming climate, which is part of why 74 percent of seasons ran below them.

What we will defend is narrower and, we think, more useful than a headline percentage. Across these five winters the published forecasts were beaten comfortably by a rule with no skill at all, 43 percent of them could not be falsified, and the gap between the almanac's self-reported accuracy and any independent number depends heavily on a city selection that readers cannot inspect.