Forecast accuracy
How FuelRadar measures predictor accuracy
No results have met the publication threshold yet. The method below explains how forecasts will be checked against later reported city averages.
Waiting for enough matched days
How we measure
How the accuracy check works
Each result starts with a saved forecast and the city average later reported for the same date and fuel grade.
Match the forecast to the reported result
FuelRadar pairs a saved city and fuel-grade forecast with the daily city average later reported for its target date.
Measure the size of each miss
The absolute error is the cents-per-litre gap between the forecast and reported average. Using the absolute gap stops high and low misses from cancelling each other out.
Use the latest 90 days
The report includes matched target dates from the latest 90-day window. Older checks are left out of the published result.
Publish only after three matches
A city and fuel grade appears after at least three forecast dates have matching reported averages. The table always shows the number of matched days beside the error.
Not enough back-test data yet
- No city and fuel grade has reached the publication threshold yet. A result appears after at least three forecast dates can be matched with later reported city averages.
Predictor accuracy FAQ
An accuracy percentage treats every miss as simply right or wrong. Average error shows how far the forecast was from the reported price in cents per litre, so a 2 c/L miss is clearly different from a 20 c/L miss.