2018 Tornadoes, High Risk, Contest Rules and Ideas

[Mark Blue]

I was thinking it would be nice to make this a contest where the winner not only received bragging rights for one year, but also gets a $100.00 gift card to Amazon, Newegg, or an online store of their choice. This would be on my dime, unless @Steve Miller wants to split it with me.

However, we need a good set of rules on how to determine the winner, plus a deadline for entries to be made. I’d also like to see a points system where going over on the tornado count or the first high risk date doesn’t disqualify the participants and a way to handle points if someone guesses no high risk days. I think adding the state might over complicate things, but am open to keeping it if it can be quantified with points. What does everyone think? Please throw out ideas on the rules if you’ve been down this road or have a bright mind for mathematics. I can split this into another thread if it results in a lot of discussion.

 

[JamesCaruso]

How about lowest number of points wins. One point for every number of tornados off from the actual total (absolute value, so whether over one or under one, you get one point); one point for every number of days off from the high risk date (five days before or after your guess, you get five points). If you guess no high risk days and there is one, you get a flat number of points added to your total. If you guess the state for extra credit, you get a flat number of points deducted. How many for each of these? Not sure, but figured I would put a few thoughts out here that others can add to [emoji848][emoji57]


Sent from my iPad using Tapatalk

 

[Quincy Vagell]

How about lowest number of points wins. One point for every number of tornados off from the actual total (absolute value, so whether over one or under one, you get one point); one point for every number of days off from the high risk date (five days before or after your guess, you get five points). If you guess no high risk days and there is one, you get a flat number of points added to your total. If you guess the state for extra credit, you get a flat number of points deducted. How many for each of these? Not sure, but figured I would put a few thoughts out here that others can add to [emoji848][emoji57]


Sent from my iPad using Tapatalk

If we do something like that, how about those who made a prediction earlier have less total points as well? Maybe something like a deduction equal to the difference between the number of days between the deadline and the entry.

For example, if the contest closes on March 31st and long range data suggests the potential for a higher end event around April 7th, there won't be nearly as much of a advantage compared to someone who has already submitted a guess.

 

[Jeff Duda]

Mark, if you're willing to foot the bill, the rules are yours to decide upon. But my suggestions are:
-compute the absolute tornado count error (so |actual tornado count - estimate|) and scale the high risk issuance so that a miss on that represents the climatological spread of seasonal tornado counts. I don't wanna take the time to compute that right now, but you would follow the methodology here to adjust the seasonal counts to account for the long term trend, then compute the standard deviation of the adjusted yearly counts. I'm guessing the standard deviation is somewhere around 150, so you could scale the high risk error to have about the same value as being off on the tornado counts by 150.

-I wouldn't worry about creating a point value for the "no high risk" category because if there is a high risk, there's no reason someone who predicted "no high risk" should win, even if they get the tornado count right on. And if there is no high risk, then there's no reason someone who predicted a high risk should win, even if they get the tornado count right on.

 

[Mark Blue]

I was going to say that if for some odd reason I won, I’d defer the gift card and give it to the person in second place. As far as tornado counts we’d have to wait until early January 2019 for the final SPC number right?

I like everyone’s ideas here. My biggest concern right now is this: is there enough time to get it up and running if the deadline is March 31st for entries? I also hope I’m not stealing the OP’s thunder by commandeering his thread and turning it into something more than he intended. What do you guys think?

 

[Jeff Duda]

As far as tornado counts we’d have to wait until early January 2019 for the final SPC number right?

Technically the final numbers are to be derived from Storm Data, which is not usually complete until about 3-6 months after the fact. So, realistically, we would have to wait until next spring. 2017 data is only complete through November at this time, if that gives an indication of how long after the fact it takes to get the final numbers.

Or you could use SPC's numbers, but be sure to specify that in the rules and give those who have already guessed a chance to change their guess to account for the fact that SPC numbers are always going to be high due to duplicated reports. SPC uses an approximation of 85% to derive final values beforehand, so that could also be a condition.

 

[Mark Blue]

I think I’m going to go ahead and award a winner this year based on some form of points as described by James above. It’ll be a test run of sorts to see how it turns out, then next year we can refine the rules based on the other feedback in this thread (or if we receive anymore during the interim) and tighten things up a bit to make it as fair as possible.

Knowing then what we don’t know now will enable us to start the contest earlier with a set of posted rules and run a more organized contest that’ll hopefully draw several more entries than we have this year. Thank you everyone for being a sounding board and giving me so much food for thought.

 

[Jeff Duda]

The prediction thread is still receiving entries. Shall we mark a deadline for guesses? Say 0600 UTC on April 1st?

 

[Mark Blue]

Good catch Jeff. Should we make it Saturday the 31st?

 

[Jeff Duda]

Good catch Jeff. Should we make it Saturday the 31st?

How about the last day of March instead?

 

[Mark Blue]

Yes the last day of March Mr. Jokester!

 

[Jeff Duda]

You started it!

 

[Mark Blue]

We’ll my High Risk prediction of 4/27/18 is looking like I totally blew it. It isn’t even May yet, so until we see some moisture making its way poleward I’m thinking we’ll be lucky to have a High Risk event this year.

Still need to finalize the contest rules as well....

 

[Jeff Duda]

There are going to be a lot of high risk busts...

I was going to put together a graphic showing the distribution of forecasts. Guess I better get on that...stay tuned.

(EDIT/ADDED):
No graphic, as sample size was rather low, but here's a text description of the forecasts.

N=37
Mean / median tornado count: 1062 / 1047
(conditional) Mean / median day of first high risk: May 6 / April 29
The condition above results from 5 predictions of "no high risk"

What is very interesting about the predictions this year is that the location of the distribution is greatly different than climatology. The average number of tornadoes in the US over the last 50ish years is ~1300. Only two predictions were for an above average count. A substantial fraction (14/37 = 38%) was for less than 1,000 tornadoes.

Also, 11 forecasts of the first high risk have already been too early, with another 6 virtually certain to also be too early. Only three forecasts called for a first high risk to occur after the end of May, and half of the forecasts for the date of the first high risk were before May 1st.

Let's see how this shakes out.

 

[Brandon Centeno]

Just wanna say that I’d like to petition to merge my Storm Track and American Wx “First High risk” guesses together given our upcoming week lmao!