Scorecard Free - Help

Scorecard Free gives you weekly updates on the performance of some of the most important weather models, for the nation as a whole. Updates run on Monday mornings and are available by 6:00 a.m. Eastern Standard Time.

For a more frequent and flexible analysis tool, ask us about Scorecard Basic. For a truly interactive analysis tool with many more capabilities, ask us about ScorecardPro.

Cities

Scorecard Free produces an aggregate statistical report for the following U.S. cities:

KABQ	Albuquerque, NM
KATL	Atlanta, GA
KAUS	Austin Intl, TX
KBNA	Nashville Intl, TN
KBOS	Boston, MA
KBUR	Burbank, CA
KBWI	Baltimore, MD
KCVG	Covington, KY
KDAY	Dayton Intl, OH
KDEN	Denver Intl, CO
KDFW	Dallas/Ft Worth, TX
KIAH	Houston/Bush, TX
KLAX	Los Angeles Intl, CA
KLGA	NYC/La Guardia, NY
KLIT	Little Rock, AR
KMCI	Kansas City Intl, MO
KMCO	Orlando Intl, FL
KMEM	Memphis Intl, TN
KOKC	Oklahoma City, OK
KOMA	Omaha, NE
KORD	Chicago/O'Hare, IL
KPHL	Philadelphia, PA
KPHX	Phoenix, AZ
KSAC	Sacramento, CA
KSEA	Seattle, WA
KSFO	San Francisco, CA

Interpretation

The weekly analysis report shows how each of several models stacks up. It does this analysis for each for three quality measures.

4.34
56

In each cell of the table, you'll find a pair of numbers. The larger number gives the quality measure and the smaller gives the size of the sample used to compute it. For each quality measure, the scorecard highlights the best model in green and the second best in yellow.

Bias (Mean Signed Error)

On average, the forecast is above or below the observation by this amount.

This measure shows how high or low the forecast is, on average, compared to the observations. Other things being equal, a model that has the lowest bias is the best model. But bias, by itself, is not enough to judge the quality of a forecast. A model with high bias but low standard deviation of signed error is better that one with low bias and high standard deviation -- because the low standard deviation means that the bias is reliable and can be subtracted out of the forecast to yield a more accurate forecast.

Mean Absolute Error (MAE)

On average, the forecast is off by this amount in either direction.

This measure shows how far off the forecast is, without regard to whether it's high or low. Other things being equal, a model with a low mean absolute error is better than a model with a high absolute error. Unlike bias, though, this measure can't be subtracted out of a forecast to produce a better forecast.

Standard Deviation of Bias

68% of the time, signed error differs from average bias by less than this amount.

This measure shows the variability of the signed error. A high value indicates that the model's forecast errors are not consistent. Other things being equal, a model with low standard deviation of signed error is better than a model with a high value. A model with a low standard deviation of bias can be bias corrected.