The National Collegiate Forecasting Contest, Page 7 of 8

At a simple city like Brownsville in the fall, the winner might have a total of 20 error points against an average of 30, while in a difficult city like Dodge City in the winter, the winner might have a total of 100 error points against an average of 120. Was the Brownsville forecaster better because he/she had fewer error points? Was the Dodge City forecaster better because he did much better than average? It's difficult to assess forecasting skill from different cities.

To allow such comparisons, and to compute an annual score, the NCWFC normalizes all the scores.
First, Penn State computes the consensus forecast each day. This is done by averaging together all the maxes, all the mins, and all the precip cats. Believe it or not, this does not produce an "average" forecast. Consensus is actually usually much better than average, so good that over the course of a year, only about 5% of all forecasters will finish above consensus.

Second, Penn State takes everybody's error point totals and computes the standard deviation over the two-week period.

Finally, all scores are normalized to a standard scale whereby the consensus forecast is exactly 80 and the standard deviation is exactly 10. Thus, if your normalized score was 80, that means you got the same number of error points as consensus. If your normalized score is 85, that means that your error points were half a standard deviation above consensus. If you can do better than 85 normalized, you'll be doing really well.

We get the official results from Penn State on the Tuesday following the end of each period. In the meantime, we keep track of the scores of about half the forecasters in the country on a daily, unofficial basis here. So you can get a sense of how you're doing compared to the rest of the country, but you won't know for sure until the following week. If you see any discrepancies between your forecasts and our scores here, tell the local chairperson (Prof. N-G).

Questions or Comments

Technical: E-mail John Fulton < jdfult@nimbus.met.tamu.edu >
Scientific: E-mail Dr. John Nielsen-Gammon. < nielsen@ariel.met.tamu.edu >