>Another interesting idea would be gathering prediction data and comparing to actual (historical) data later, that should find most precise weather forecast service.
I've been interested in doing this for awhile, but I've had trouble finding a data source of past predictions from their sources. It's also a frustrating exercise to try to search for online.
If anyone has data sources or additional context/keywords available, I'd be interested!
For a couple years I ran an automated script which grabbed the 10 day forecast from a weather site and was used for energy forecasting at a manufacturer. I archived the data after scraping it. As you'd guess, the variability of the weather is notably higher 10 days out, but the forecasts were remarkably accurate on a short term.