The Weather Forecast Challenge
You're the forecaster. The line shows the temperature up to now. Pick how far ahead to forecast, drag your prediction up or down, and lock it in. Then watch what the weather actually does — and see how close you got.
Make a forecast
Drag up or down on the chart to set your predicted temperature.
Result
Near is easy, far is hopeless
Forecast just 1–2 hours ahead and you'll usually nail it — the line keeps doing what it was doing. Forecast 18–24 hours ahead and you're basically guessing.
On your record below, the dots stay low on the left and scatter high on the right. That rising wall is your prediction horizon.
Why forecasts have an expiry date
This "weather" follows fixed rules — it's the same chaotic system as the Butterfly Effect lab. Tiny things you can't see grow until the forecast falls apart.
Real meteorologists hit the exact same wall. That's why you trust tomorrow's forecast far more than the one for two weeks out.
For teachers & grown-ups
The "temperature" is the x-coordinate of the Lorenz system (σ = 10, ρ = 28, β = 8/3), sampled each "hour" and shifted into a comfortable °C range — the same chaotic flow used in the Butterfly Effect lab. Because a human forecaster extrapolates the visible trend, accuracy is high while the trajectory stays on one wing and collapses when it switches wings. The scatter of error versus lead time makes the forecast-skill decay visible: error stays small out to roughly one Lyapunov time and then saturates at the climatological spread (guessing). A good discussion prompt: what would make long forecasts better — a faster computer, or more precise thermometers? (Neither beats the exponential growth of error for long; both only push the wall out a little.) Compare a student's forecast against "persistence" (assume it stays the same) to introduce the idea of forecast skill versus a baseline.