Before the annual crop of hurricane season forecasts, ace AIR blogger Scott Stransky organizes a fun competition within AIR. The first part challenges staff to estimate the insured hurricane losses that will be incurred in the U.S. during the upcoming season. The second dares us to predict the total of named tropical storms that will form in the North Atlantic basin, how many of those will become hurricanes, the count of major hurricanes (Saffir-Simpson Category 3 and higher), and the number of hurricanes that will make landfall in the U.S.
This year there were 131 participants chasing seven modest prizes using techniques ranging (so I am told) from running complex computer simulations to consulting psychics. I won the Storm Counts contest by guessing predicting with uncanny accuracy that there would be 8 named storms, 5 hurricanes, 2 major hurricanes, and 1 U.S. landfall. The fact that 6 hurricanes actually formed cost me one error point, but still left me with a marginal lead. Second and third places had to be decided by a tie breaker.
Overall, our estimates varied widely: named storms (2 - 22),hurricanes (1 - 15), major hurricanes (1 - 20; note that this implies that at least one person predicted more major hurricanes than hurricanes), and U.S. landfalling storms
(0 - 7).
Scott suggested that I should blog about how I had come up with such an accurate prediction-but of course mine wasn't a coldly calculated forecast based on ample and accurate data. Rather, it was a wild guess based on a vague hunch-like feeling that maybe 2014 just possibly might be another quiet Atlantic hurricane season-perhaps.
Obviously, given the number of named storms there would be fewer hurricanes, an even smaller number of major hurricanes, and fewer still U.S. landfalling storms. I simply picked numbers that fit this pattern, and the ones I chose closely matched what transpired. I got lucky!
I didn't do at all well in the U.S. Hurricane Loss Contest however. This year we had an occurrence that was not accounted for in Scott's rules. Hurricane Arthur made landfall in North Carolina as a Category 2 storm, but neither ISO's Property Claim Services®nor AIR released loss estimates. Although the losses were low, they were certainly non-zero and another official source had to be found. The National Climatic Data Center report on Arthur, which suggested losses of USD 2,346,000, came to the rescue.
AIR's winner estimated USD 500, the lowest non-zero loss, and demonstrating an admirable strategy. The figure I came up with-USD 600 million-was way off the mark, but much closer than some of my colleagues managed. While many of us were in the ballpark (or at least, in the right neighborhood), overall the loss predictions ranged from USD 0 to USD 53,000,000,000! Three colleagues tied for second place even after the tie breaker-please don't ask me how the rules work!
The hurricane contest is a lot of fun, but although 41 participants were better than the median for BOTH contests, most of us did not come up with very accurate predictions. Hurricane seasons are notoriously difficult to forecast-particularly when you have limited data to work with.
Would you like to participate in our 2015 contest?