2018 Tornadoes, High Risk, Contest Rules and Ideas

Trade Stocks for Free with Robinhood
Leverage your knowledge of weather to beat the stock market with free trading on Robinhood! Sign up today, get free stock!

Mark Blue

Staff member
Supporter
Feb 19, 2007
2,793
325
21
Colorado
Thank you Jeff for tracking this and narrowing down the potential winners. I really appreciate what you’re doing here as it’s helping me out a bunch!
 

Jeff Duda

Resident meteorological expert
Staff member
Oct 7, 2008
3,015
1,534
21
Westminster, CO
www.meteor.iastate.edu
Mark, here are the figures I claimed I could produce back in March, but only just now got around to doing:

This one adjusts yearly tornado counts for the increase over time due to better reporting:
tornado_counts.png

This is just a histogram form of the above data:
tornado_distribution.png
confirming that the standard deviation of the adjusted yearly counts is ~200, which would have been my suggestion for a tornado count error that would be comparable to being off on the "high risk" category (if you chose to score it that way). So conceivably, someone who went with a high risk could win by being much closer to the actual tornado count than someone who went with "no high risk" but was significantly off on the count.

It's up to you how you want to weight that, though.
 
  • Like
Reactions: Jeff House

Mark Blue

Staff member
Supporter
Feb 19, 2007
2,793
325
21
Colorado
Edited...

After reading yesterday’s post again if there are 3 people who guessed no high risk and it’s all about count now then the 3 who are left should be the last ones standing for the prize based on count. Unless a high risk pops up between now and the end of the year, which would be next to impossible.
 

Jeff Duda

Resident meteorological expert
Staff member
Oct 7, 2008
3,015
1,534
21
Westminster, CO
www.meteor.iastate.edu
After reading yesterday’s post again if there are 3 people who guessed no high risk and it’s all about count now then the 3 who are left should be the last ones standing for the prize based on count. Unless a high risk pops up between now and the end of the year, which would be next to impossible.
I may have masked this in my previous posts, but my point was to ask if you wanted to consider the fact that there are some contestants who did go with a high risk, but who also predicted tornado counts that are very close to the current actual counts. Thus someone who went with a high risk may end up being closer in total tornado count than the closest of the three who went with "no high risk. That will be the harder choice to distinguish between.

In other words, I don't think that the winner must necessarily be someone who also went with "no high risk".
 

Mark Blue

Staff member
Supporter
Feb 19, 2007
2,793
325
21
Colorado
If someone went high risk and has a close tornado count, we should probably deduct a standard number of points IF the year ends with no high risks being issued.

The contestant with the lowest score would be the winner or so it would seem.

For tornado counts I believe we should just count the number of days from the final number, regardless of whether the guess is over or under, so we don’t have negative and positive numbers. I’m really leaving a lot up to you because of your mathematics background. Did you have ideas that are way different than the above?
 

Jeff Duda

Resident meteorological expert
Staff member
Oct 7, 2008
3,015
1,534
21
Westminster, CO
www.meteor.iastate.edu
I had suggested scaling the tornado count error by the spread of trend-adjusted annual tornado reports, which, as I reported a few posts back, is 203. Thus, take the absolute value of the tornado count error and divide by 203. From that value, add 1.0 if the person was wrong about whether or not there would be a high risk, and add 0.0 if the person was right about it. Lowest score wins.

Usage:
Suppose "no high risk" verifies. Under this scoring method, it's not guaranteed that someone who went with "no high risk" would win. If the tornado count ends up significantly increasing to the point where the person who went with "no high risk" who was closest on tornado count (among all who forecasted "no high risk") ended up off by more than 203 tornadoes than someone who went with a high risk, then the person with the closer tornado count who did go with a high risk would end up with a lower score than the person who went with "no high risk".

Example:
Say "no high risk" verifies and the tornado count is 1450. Say person A says "no high risk" and 1000 tornadoes; person B says "high risk on XXX day" and 1400 tornadoes. In this case, person B would end up with a lower score than person A.

Bottom line:
This scoring method unfortunately does not account for how to differentiate between people with different days of "first high risk" in the event that "no high risk" verifies. However, given the current values, I don't think there is going to be a problem justifying giving the win to one of the 5 people who went with "no high risk" (again, assuming "no high risk" actually verifies). Even if the final count ends up way higher than 900 or so, we also had two additional contestants in the "no high risk" category in the 1100-1200 range, which should allow a person who went with "no high risk" to beat someone who went with one.

Alternative scenario:
In the unlikely event that a high risk actually does occur, there was a contestant who went with a November high risk date, and they were pretty alone out there. Their tornado count would probably still end up a bit high, but they would probably end up close enough to justify giving it to them instead of one of the "no high risk"ers.

Bottom-bottom line:
There are realistically only 6 people who have a chance at winning this at this point. Some crazy shit would have to go down in the next 50 days for one of those 6 not to be the winner.
 

Mark Blue

Staff member
Supporter
Feb 19, 2007
2,793
325
21
Colorado
There are a couple of guesses that are very close number-wise, but the persons guessed a high risk date when there was none. How does it work out @Jeff Duda when you run those close guesses through your formula? I just want to be sure everyone gets a fair shot at the gift card.
 

Jeff Duda

Resident meteorological expert
Staff member
Oct 7, 2008
3,015
1,534
21
Westminster, CO
www.meteor.iastate.edu
There are a couple of guesses that are very close number-wise, but the persons guessed a high risk date when there was none. How does it work out @Jeff Duda when you run those close guesses through your formula? I just want to be sure everyone gets a fair shot at the gift card.
With SPC showing 991 as the inflation adjusted running number, it seems clear that whoever picked 967 is or will be the winner. That is unless there’s someone with a closer number to 991 and no high risk. What do you think?
Given two people guessed 950 and 967, I am not comfortable declaring the 967 as the winning tornado count yet. We really should wait a few months (especially once the shutdown ends) and get a more accurate "confirmed" count before awarding the victor. It's possible the 991 will not be the final number.
 

Mark Blue

Staff member
Supporter
Feb 19, 2007
2,793
325
21
Colorado
Thanks Jeff. I saw the message on their website about the closure so it wouldn’t hurt to wait a little while longer. I hope this closure is about to end as it’s sad so many federal government employees aren’t getting paid.