The Experiment:  For two weeks, we compared the five-day forecasts of four Cleveland, Ohio area meteorologists to the actual weather. We used the chief meteorologists from the four Cleveland, Ohio television networks (Dick Goddard, Bruce Kalinowski, Mark Nolan, and Brad Sussman) as our contestants. To compare the accuracy of their forecasts, we used the following process:

1. On Sunday evenings, we used the television stations' Web sites to acquire the forecast highs for the next five days.
2. At the end of the week, we used the data on http://www.accuweather.com/ to obtain the actual recorded high temperatures for those five days.
3. We then used this data to calculate the forecasters' daily error and absolute error.
4. Finally, we whipped up a couple cool graphs in Excel.

The Results:  As you can see from the chart below, there was no runaway winner. The forecasters main mistake was underestimating the high temperature, which I'm sure most Clevelanders like us would forgive if it means unexpectedly warm spring weather!

Mark Nolan had the lowest error and absolute error scores, but of course there is some controversy over these results. Brad Sussman had the worst scores, but by only 0.2 degrees. If we were being scientifically rigorous, we would bet there is probably no stastically significant difference in the results, although interesting trends do begin to appear. (Of course, if you decide to pull out your Introduction to Probability and Statistics book and do the math, feel free to send us the results.) :-)

Individual Results
Dick Goddard   Bruce Kalinowski   Mark Nolan   Brad Sussman

First, out of laziness. Forecasts for high temperatures and actual high temperatures are easy data to collect. Second, out of fairness. Temperatures are cold hard facts and allow for easy, understandable calculations on how close a forecast came to the truth. Other measures are harder to pinpoint; one man's partly cloudy day is another's mostly sunny day.

Mark Nolan had the lowest average error (over- and under- estimates all rolled into one) and the lowest average absolute error (absolute number of degrees away from the real high). However, his score was greatly aided by the fact the Channel 3 Web site was not updated in time to allow for data for March 15th. March 15th was the worst day for all our other forecasters, and presumably would have been for Mark, too.

We thought so too, but we included the "Technology" section because we were curious if the "fancy-sounding-ness" of the station's weather technology would have an effect on the results. Unfortunately, "fancy-sounding-ness" had a good correlation with accuracy. (Triple Doppler best, Dual Doppler OK, and single Dopplers trailing.)

We fear this will have the very unfortunate consequence of endless commercials for new Quadrillion Dopplers and Infinity Plus One Dopplers as the stations continue to try to outdo each other.

Sure. You can view or download the Excel file.

Feel free to use the Excel file from the link above to pit the forecasters in your town against each other. If you've got the time, try to keep it going for longer than two weeks so we can see if any significant differences emerge. If you do it, be sure to tell us about it!