How far ahead is the weather forecast accurate?
In short: trust the temperature for about three days and the rain for about one. A rain call for tomorrow comes true a bit over half the time; one made more than three days out comes true fewer than three times in ten.
- Temperature barely slips for a day. Our five forecasters averaged 0.8°C out for both the next hour and the next 24.
- It drifts after that: 1.0°C out at one to three days, 1.4°C at three to ten.
- Rain calls fade fastest. 56% came true within a day, 47% at one to three days, 28% beyond three days.
- Dry calls stay right about 93% of the time at every range, mostly because nine hours in ten are dry anyway.
How many days ahead can you trust a weather forecast?
About three days for temperature and wind, and about a day for rain. That's from marking 2.4 million forecasts from five forecasters (the Met Office, Apple, YR, DWD and Tomorrow.io) against what the weather actually did, from June to October 2026, sorted by how far ahead each one was looking.
Here's the average across all five, for each range. "Rain call came true" means: when the forecast said rain, did it actually rain?
| How far ahead | Temperature miss | Wind miss | Rain call came true | Sky right |
|---|---|---|---|---|
| Next hour | 0.8°C | 3.0 kph | 53% | 67% |
| 1 to 3 hours | 0.8°C | 3.0 kph | 55% | 67% |
| Up to 24 hours | 0.8°C | 3.0 kph | 56% | 67% |
| 1 to 3 days | 1.0°C | 3.3 kph | 47% | 63% |
| 3 to 10 days | 1.4°C | 3.9 kph | 28% | 56% |
How accurate is tomorrow's weather forecast?
Tomorrow's forecast is nearly as good as the next hour's. Temperature was 0.8°C out on average, wind 3 kph, and rain calls came true 56% of the time, which is as good as rain calls get. If you're planning tomorrow's barbecue, the forecast is worth listening to.
How accurate is a 3-day forecast?
A forecast for two or three days away is still useful, but rain calls drop to slightly worse than a coin toss. Temperature slips to about a degree out and the sky gets harder to call. It's fine for deciding which day looks best, and less good for deciding which hour.
How accurate is a 10-day forecast?
A 10-day forecast is good for the general picture, not the detail. Temperatures were 1.4°C out on average, which is still handy for "will it be a coat weekend?". Rain is where it falls apart: a rain call made more than three days ahead came true only 28% of the time, so a wet icon next Saturday is a maybe, not a washout.
The forecasters also behave differently that far out. Apple almost stops calling rain beyond three days and saw only 4% of the rainy hours coming. YR goes the other way, catching 63% of them but crying wolf four times in five. Neither tells you much about a specific hour next week.
Are dry forecasts reliable further ahead?
Yes, a dry call was right about 93% of the time whether it was made an hour ahead or a week ahead. Before you get too excited, that's partly because about nine hours in ten in Britain are dry anyway, so "dry" is the safe bet. Still, it means a dry forecast for the weekend is worth a lot more than a wet one.
Why isn't the next hour's rain forecast better?
The hourly forecasts in this test are weather models, and models are poor at pinning a shower to one hour in one street, even an hour ahead. For the next hour, radar and minute-by-minute nowcasts do much better: in our own tests over a few wet days in October, radar tracking caught 93% of the rain. That's why Weather for Dads uses a separate next-hour panel rather than the hourly forecast.
Which forecast holds up best further ahead?
It depends what you're checking. YR was closest on temperature for anything up to three days ahead, with Apple a whisker ahead of it beyond that. For rain beyond three days, the Met Office had the best score, and it's the forecaster we'd lean on for a long-range wet-or-dry call. The full breakdown by range, for every forecaster, is on our accuracy page, and we compared them overall in which UK weather forecast is most accurate.
How should you use the forecast for a weekend plan?
Look early for the trend, then decide late. A week out, check whether it looks warm or cold and whether it's mostly dry. Two or three days out, pick the better day. The day before, pick the hours. And whatever the range, treat a rain icon as a "maybe": what a chance of rain actually means explains why.
Sources
Our figures: Weather for Dads accuracy page. Five forecasters (Met Office, Apple, YR, DWD, Tomorrow.io), 12 June to 10 October 2026, marked against Met Office observing stations. Rain means 0.2mm or more in the hour. Figures in the table are the average of the five; "sky right" is how often the forecast sky (sunny, cloudy, rain and so on) matched what was observed.
Radar test: Met Office radar composite, from the Met Office open data on Amazon Web Services, tested 8 to 10 October 2026 against Environment Agency rain gauges. Early figures over about 50 wet hours.
Five forecasts blended into one plain answer, with every one of them marked in public against what actually happened. Free, no account, no ads.
On Android or a computer, open the app and tap "Add to phone Home Screen": it works like any other app. A Google Play version is on its way.