I suppose this is like a bank robber who robs a bank, then promptly asks to deposit their newfound loot into their account at the same bank.

I've always wondered, but never had the time nor determiniation to find out, whether weather forecasts are more or less accurate than coin tosses and dice. I've always wanted to run an experiment similar to this :

Take the 5 day temp forecast from a national weather site (NWS/weather.com, etc) and write the high / lows on one row.

Then, flip a coin and roll a single die.
Start with yesterday's high and low.

If you toss "heads" then add the number rolled on the die to the high temp for that day. If you toss "tails" then subtract. Do this for high and low temps for each of the five days.

Repeat this routine for a while, logging actual temps, then, after a certain amount of time. Compare your "forecasts" to the NWS / weather.com forecasts.

I've always wondered which would be more accurate.

Maybe this guy would run that test! ;)