I was looking back at the predictions for the 2019 North Atlantic tropical cyclone season in Wikipedia the other day, and although they produced a good looking table of all the various predictions, they didn’t take it to the next level and score each of them. Perhaps that’s not the way they do things at Wikipedia, but in the xmetman blog we do things slightly differently, so now that the HURDAT2 data is in for 2019 I decided to come up with my own scoring system. courtesy of Google sheets and a bit of plagiarism.
In the table below I have applied a weighting factor to the difference between the actual and forecast totals for the three forecast categories. The weighting is a point for each named storm, three points for a hurricane and five points for a major hurricane. So if you deduct the forecast from the actual and the multiply by this weighting you end up with a total score. A negative score means that you have been unduly pessimistic and a positive score overly optimistic about tropical storm activity, a perfect score is zero. It’s a bit rough and ready and perhaps I should have added the absolute values, but what the hell.
Closest to that perfect score of the seven forecast sources for 2019 was TWC with a score of -1. TWC is an acronym for The Weather Company if you’re wondering, I also notice it’s also owned by IBM which in itself is interesting. The Met Office also came up with a score of -1 which is pretty good even if they do think the hurricane season ends on the 1st of November (well they would have to be different wouldn’t they). As you can see some forecasts are done multiple times through the season, TSR (Tropical Storm Risk) for example, and sometime the later ones are less accurate than the earlier ones! I don’t think my money will be on them for this upcoming 2020 season.