AN INTRODUCTION TO THE
SAINT CLOUD STATE FORECASTING CONTEST

Play the SCSU Forecasting Contest!

• Spring Semester 2010
  • The Second City Is... (Deadline: Wednesday, Dec 9 11 PM)
  • EAS 485 Students Must Forecast for Both Cities
  • Participation when not required is a good idea to gain experience.
• First Day to Forecast: Monday, January 11
• If you've never played the game before, you must sign-in when you log on the first time.
  • Join EASGENERAL class if you're taking EAS 380 or beyond.

Updated: Tuesday, January 5, 2010 11:51 AM

A contest to predict high and low temperatures and the amount of precipitation as reported at the Saint Cloud airport is held on every day that classes are in session during the quarter. The game is played as follows:

TIME PERIODS: There are 4 time periods that you forecast for on any given day. Each time period is 12 hours long. For example, on Monday, you would forecast for Monday night, Tuesday, Tuesday night, and Wednesday as shown on the web page. You must make all forecasts in each time period to have that forecast count. The actual cut offs for the time periods are given in Greenwich Mean Time (GMT). They are noon (1200 GMT or 12Z) and midnight (0000 GMT or 00Z) in Greenwich time. Since Saint Cloud is in the Central North America time zone, we are 6 hours behind Greenwich, England when we are on standard time and 5 hours behind when we are on daylight time. So, the forecasting time periods for a Monday forecast would be as shown below:

Note that, for the second city game, the time periods are as follows:

TEMPERATURES:

• For each time period, you must forecast a temperature in degrees Fahrenheit (°F).
• If the period is a nighttime period, forecast the lowest temperature that the St. Cloud Municipal Airport will reach between 00Z and 12Z.
• If the period is a daytime period, forecast the highest temperature that the St. Cloud Municipal Airport will reach between 12Z and 00Z.
• Because of the way the time periods are set, you can never have a daytime high temperature lower than the previous night's low.
  • For example, if the temperature at St. Cloud Airport fell a degree an hour from 20°F at midnight CST on Tuesday morning until 9 PM CST Tuesday evening, the temperature at 6 AM CST on Tuesday would be 14°F. So, the Monday night low would be 14°F. Now, if the temperature kept falling a degree an hour until 6 PM CST, it would be 2°F at 6 PM CST. However, the highest temperature between during the daytime period on Tuesday is the highest temperature between 6 AM and 6 PM CST. This would again be the 6 AM temperature of 14°F. So, it is impossible to get a daytime high lower than the previous night's low. Similarly, it is impossible to get a low temperature to be higher than the previous day's high.

PROBABILITY OF PRECIPITATION: There are two precipitation probability games to be played in each time period. The first is the probability of measurable precipitation. Measurable precipitation is defined as precipitation which amounts to at least .01 inch. A trace of precipitation does not count. If the precipitation takes the form of snow, the precipitation is the melted water equivalent amount, so this category verifies during a snowstorm. Note that the verification is taken at Saint Cloud State University for a snow event. The forecast is made in tens of percent.

• So, if you think that there is a 10% chance of measurable precipitation, select a "1" in the column PROB OF PRECIP of the category.
• If you think there is a 20% chance, write a "2", etc.
• If you don't think there is any chance, write a "0".
• If you think there is a 100% chance, write "10".
• If St. Cloud airport does get measurable precipitation, the verification is "10". If St. Cloud gets either a trace or stays dry, then verification is "0". You might think then that that it would be advantageous to write only 0's and 10's, but the error points in the probability game are squared, so be very careful (see scoring for more details).

PROBABILITY OF AT LEAST .25 INCH: The probability of a quarter inch (PROB>.25") game is played the same way as the probability of measurable precipitation, except that you are determining the chance that the airport will report at least 1/4 inch (0.25") of liquid water. Again, in a snowstorm, this category verfies as the melted precipitation, as measured at Saint Cloud State University.

Important: If the airport reports at least 1/4 inch, there must be at least .01 inch and both probability games verify as a "10". Therefore, you cannot enter a probability on the .01 game that is lower than the probability of .25 inch!

PRECIPITATION AMOUNT (P-AMT) GAME: In this game you are attempting to forecast the amount of liquid water that will be measured at the St. Cloud airport with an error of .05". To do this, you write down a precipitation amount category. The categories are shown below.

Note that there are many possibilities for precipitation amount. Each category is .05" wide and the highest amount is the category number times .05". We have had a category 89 verification (4.46 - 4.50 inches) in Saint Cloud. Again, in snow situations, it is the liquid water equivalent that is used to determine the category.

Note also, for categories 1-5, the Probability of Precipitation is "10" and the Probability of Quarter-Inch is "0" for all verifications except .25 inch. For all categories of 6 or high, both the Probability of Precipitation and the Probability of Quarter-Inch verifies as "10."

SNOWFALL GAME: In this game, you are attempting to predict the amount of new snow that will fall at Saint Cloud State Univerisity in whole numbers of inches during a given time period. The verification amounts are rounded to the nearest inch. Therefore, if you forecasted 2 inches to fall on Monday night and SCSU reported 2.4 inches, then the 2.4 would be rounded to 2 and you would have hit it on the nose. This also means that snowfall verifies as zero unless SCSU receives at least 0.5". Zero is a legitimate forecast in this game.Also, note that, for a snowstorm, you should have a number above zero in both the P-AMOUNT and SNOWFALL games if you expect at least 0.5 inch of accumulation and at least .01 inch measurable (usually a solid dusting).

ENTERING FORECASTS:

1. The forecast game runs on all class days from Monday, September 8, 2003, through Friday, December 12, 2003. This means that there are 71 forecast days in Fall Semester 2003.
2. You must participate on 60% of the total available forecast days to meet the requirements for EAS 380 (Saint Cloud only) and EAS 385 (both Saint Cloud and Watertown, NY), to be listed in the composite standings, or to qualify for Bob Weisman's Forecast Challenge. Each week's standings the minimum number of forecasts needed to qualify on the total standings. For Fall Semester 2003, the minimum number of forecasts will be 42 forecasts.
3. No forecast counts unless you push the "Submit" buttom.
4. All forecasts must be submitted by 4:00 PM Monday through Friday in order to count in the game.
5. Only forecasts for that given day will be allowed.You cannot fill out Friday's forecasts and submit it on Thursday with Thursday's forecast.
6. You must fill out every forecast for all 4 time periods (2 time periods for the second city) for your forecast to count.

HOW THE GAME IS SCORED:

The forecast standings are expressed as a percentage improvement over a "consensus" forecast. The consensus forecast is the average forecast of all EAS department forecasters who played that day. Therefore, the idea of the game is to make a better forecast than "consensus". This method helps to balance the difference between "easy" forecasting days (when everyone is very close to the actual temperature) and "tough" forecasting days (when most people are way off). An illustration is given below.

Anderson's Forecast

Weisman's Forecast (a.k.a. The Evil Empire):

If Anderson and Weisman are the only 2 players in the game, then consensus would be the average of the 2 forecasts and would appear as follows:

CONSENSUS FORECAST

Let's say that what actually happened was that no precipitation was reported and Monday night's low was 14, Monday's high was 37, Tuesday night's low was 20, and Wednesday's high was 36. So, the verification is:

The forecast game, if you used the predict history menu for this date, would look like this for Anderson:

On Monday night, Anderson missed the low by 2 degrees, but consensus missed it by 3 degrees. So, Anderson gains a degree error point on consensus. Anderson was off by 5 degrees on his Tuesday high while consensus missed by 3 degrees there, so he loses two degree error points on that forecast. Similarly, Anderson gets 5 error points on Tuesday night's low while consensus gets 6 and Anderson gets 2 error points on Wednesday's high while consensus gets 2. So, for Monday's temperature forecasts, Anderson got a total of 2+5+5+2 = 14 error points while consensus got 3+3+6+2 = 14 error points. The percentage improvement on consensus for the temperature and precipitation amount games are based on the following formula.

S = Total consensus error points - total forecaster error points X 100‰
                      Total consensus error points

So, Anderson's temperature score for Monday would be

(14-14)/14 = 0.0%

For the probability game, Anderson only picked up error points in the Probability of .01 inch game. Note, however, that these errors points are squared. So, Anderson picked up 1 POP error points for Monday night and 4 POP error points for Tuesday for a total of 5. Consensus also received 4 POP error points for Monday night and 4 more POP error points for Tuesday for a total of 8.

In the probability of .25 inch game, Anderson had zero error points, but consensus (thanks to Weisman's rotten forecasting) has 1 error point.

The POP error points of both POP games are combined then turned into a percentage using the formula above. So, Anderson's POP score is: (8-5)/8 = 37.5% and his POP.25 score is: (1-0)/1 = +100.0%. The error points for both games are combined into a total POP score. So, Anderson's all POP score is: (9-5)/9 = +44.44%

Another percentage score is made from the P-Amount games (P-Amount and Snowfall). In this case, Anderson again has zero error points, but consensus has 2 error points. So, the total P-amount score is: (2-0)/2 = +100.0%.

The total game score for each forecast is:

Total Score = .50(Temperature Score) + .25(POP Score) + .25(P-Amount Score)

Running totals of error points are used to make the total standings, i.e., we don't average the daily scores...we compute the total score through all forecasts using the running error point totals. So, for the day, Anderson has earned the incredible score of +36.11%. Here's another example. In practice, it is very hard to beat consensus.

Time Periods	Temperature Rules	Probability of Precipitation		Probability of .25 inch	Precipitation Amount
Snowfall	Entering Forecasts	How the Game is Scored		Bob Weisman's Forecast Challenge	Top of Page
Spring 2010 Second Forecast City Selection	Meal Mooching	Saint Cloud Normals and Records		Tuscon Int'l AP Climatology from NWS Western Region	Arizona Surface Reports from MesoWest Project
				NWS Tuscon	Records for Tuscon from NWS Tuscon
Return to:	SCSU Forecast Web Site	Forecast Links		Current Weather Links
	EAS 380 Site	EAS 381 Site	EAS 385 Site	EAS 485 Site	Handbook

BOB WEISMAN'S FORECAST CHALLENGE: Anyone who makes the minimum number of forecasts in a given semester and beats Bob Weisman in the final standings at the end of the semester gets treated to a free meal. Recent semesters have proven that Bob's forecasting is the only thing balancing out the weak economy.

Spring 2010 Second City Selection

Last Updated: Tuesday, January 5, 2010 11:51 AM

Send questions or comments to Bob Weisman