Category: technology

The forecast said 30 percent chance of rain, it poured all afternoon, and you concluded the app is useless. The app, however, was probably right — and understanding why says a lot about how we misread probabilities.

A 30 percent chance of rain does not mean light rain, or rain for 30 percent of the day. It typically means that in 30 percent of situations exactly like this one, measurable rain fell at your location. Forecasters verify these numbers relentlessly: when major services say 30 percent, it rains almost exactly three times out of ten. The forecast is calibrated; our intuition is not.

The bigger problem is resolution. Summer storms can drench one neighborhood and skip the next, while an app shows a single icon for an entire city. That icon compresses a complex probability field into a cartoon sun peeking behind a cloud — and the compression, not the science, is what fails.

Meteorology has quietly become one of the great prediction success stories: a modern five-day forecast is as accurate as a one-day forecast was in 1980, thanks to satellites, ensemble models and vast computation.

The practical fix is simple. Check the hourly probability curve and the radar, not the daily icon. And when the app says 30 percent and you get soaked, remember: improbable is not the same as impossible — it is just Tuesday, three times out of ten.