Rules for the Forecast Contest: Tropics Edition

[Mallow]

Hi all!

It's that time of year again! Time for the annual tropical forecast contest. I have a set of rules to share with everybody. There are two major parts to the contest.

1. Monthly forecast for the number of tropical systems in the following month

Give a forecast for the number of storms which reach at least tropical storm strength, at least hurricane strength, and major hurricane strength, for the following month. For example, I might give a forecast of 5.3 storms, 3.1 hurricanes, and 2.0 major hurricanes in September, and the forecast would be made at the end of August. A guess that makes no logical sense (i.e. 3 storms, 4 canes, and 5 major hurricanes) will not be accepted. A verified hurricane counts as both a hurricane and a tropical storm in the monthly counts, and a major hurricane counts as all three. For strategic reasons, you can give a forecast to the nearest tenth. Obviously, verification will be in whole numbers, but tenths can still be statistically advantageous in cases where you're split between two possible whole-number guesses.

Each variable during each month, there is a max amount of points available, given in table 2 at the end of this post. The points for the monthly numbers are based on a total of 1,000 available for the whole contest (7 months), how many storms are average in the given month, and what the standard deviation is that month. As you'll see, the standard deviation actually comes into play twice for the monthly forecasts.

The points you are awarded for a given variable (storms, canes, or majors) in a given month are based on four things: your guess, the actual value, the standard deviation for the variable that month (see table 1 at bottom), and the points available for the variable that month (see table 2 at bottom). The calculation is as follows...

Find your absolute z-score by taking the absolute value of the difference between your guess and the actual, and dividing it by the standard deviation. Use the z-score to find a two-tailed percentile from a normal curve. Multiply the percentile by the total points available. That is your score for that variable for that month.

Basically, the better your forecast, the more points you get. And the more "important" forecasts (the heart of the tropical season) are worth more points.

You don't have to play every month to qualify... you can play any number of months. All that's important is your total score.

I have posed a question in the poll for people to answer. I like both methods about equally well.

I'll get to the second part of the contest (and the tables) later, I gotta run!

EDIT: You can submit a new forecast (in a new post) any time up to the deadline. I'll use your latest forecast issued before the deadline.

Late forecasts get docked 25% for the first 24 hours, 50% for the second 24 hours, and 75% for the third 24 hours. Any forecast over 72 hours late is not considered valid.

Also, FYI, for major hurricanes in June, and hurricanes and major hurricanes in December, since there were no such events in the thirty year climatological period of record, a standard deviation is undefined... so to make calculations possible, I am using 0.1

[phil882]

I picked option one, because option two would make it easy for folks to hedge their bets on developing systems if there is a TD the day before the month starts.

[Srain]

Option 1 was my choice. The 'official' designation would be the two digit number when the storm is classified (month) a depression forms. But that's just me.

[Mallow]

I picked option one, because option two would make it easy for folks to hedge their bets on developing systems if there is a TD the day before the month starts.

True, but the same argument could be made for developing invests being upgraded to depressions.

[Mallow]

Option 1 was my choice. The 'official' designation would be the two digit number when the storm is classified (month) a depression forms. But that's just me.

That sounds like as good a reason as any.

[kab2791]

Option 2. If a TD develops Aug 31 and becomes a TS Sept 1, this shouldn't count toward an extra TS for Aug since it isn't actively a TS in Aug.

[Mallow]

2. Monthly forecast for the total number of tropical systems during the season

Give a forecast for the number of storms which reach at least tropical storm strength, at least hurricane strength, and major hurricane strength, for the entire season. For example, I might give a forecast of 12.8 storms, 8.1 canes, and 4.7 majors for the season, and the forecast might be made at the end of September, for example. A guess that makes no logical sense (i.e. 3 storms, 4 canes, and 5 major hurricanes) will not be accepted. A verified hurricane counts as both a hurricane and a tropical storm in the monthly counts, and a major hurricane counts as all three. For strategic reasons, you can give a forecast to the nearest tenth. Obviously, verification will be in whole numbers, but tenths can still be statistically advantageous in cases where you're split between two possible whole-number guesses.

Each variable during each month, there is a max amount of points available, given in table 2 at the end of this post. The points for the seasonal numbers are based on a total of 1,000 available for the whole contest (7 months), and the average number of storms left in the season. The fewer average storms left, the fewer points the month is worth.

The points you are awarded for a given variable (storms, canes, or majors) in a given month are based on four things: your guess, the actual value, the standard deviation for the variable that month (see table 1 at bottom), and the points available for the variable that month (see table 2 at bottom). The calculation is as follows...

Find your absolute z-score by taking the absolute value of the difference between your guess and the actual, and dividing it by the standard deviation. Use the z-score to find a two-tailed percentile from a normal curve. Multiply the percentile by the total points available. That is your score for that variable for that month.

Basically, the better your forecast, the more points you get. And the earlier in the year a seasonal forecast is made, the more points it's worth.

[Mallow]

Option 2. If a TD develops Aug 31 and becomes a TS Sept 1, this shouldn't count toward an extra TS for Aug since it isn't actively a TS in Aug.

Also a good point, and that's why I had it that way last year.

[wxmx]

Option 1... with option 2 in case a TD forms prior to 2012 but it's upgraded to a named storm in 2012, it's officially counted as a 2011 storm, but wouldn't get accounted for this contest. It has never happened, but Alice 1954 and Zeta 2005 came close, so it's possible.

[wxmx]

True, but the same argument could be made for developing invests being upgraded to depressions.

But TDs are not at play in this contest, so there's still one more "step" to score in this scenario vs the TD->TS one.

[wxmx]

Also a good point, and that's why I had it that way last year.

But H and MH did... better have it consistent at all levels...IMO.

[Mallow]

TABLE 1

TABLE 2

[Mallow]

Looks like the people have spoken. Option 1 it is!

The storm's "month" will be the month in which it was first declared a depression. There are pros and cons to this method.