A leaderboard for amateur forecasts

[category_high_five]

Hi everyone! Long time lurker and hurricane enthusiast since I used to sit in front of the Weather Channel all day when I was four.

I've wanted for a long time to be able to actually try to test my forecasts against the experts. And I finally have enough amateur developer skills to actually build it! It's still a work in progress that I wanted to hack away at for another week, but with a pretty exciting storm in Beryl spinning out there it seems worth posting now even as a work-in-progress. It's a very minimalist hacky design right now built on Airtable, but it works to capture forecasts and I'm in the process of getting the scoring down. Any eligible forecast submitted now will be able to get a score retroactively later.

Working right now:
- Submit forecasts
- Best Track updates and storms added automatically
- Display names (and automatic random names if you don't submit one)
- Averaging for track and forecast scores (but these depend on the NYI track point scores)

In progress:
- Scoring track points vs the NHC best track
- Adding the official NHC forecasts to scoring, with the same formula as user forecasts, so you can compare with the pros
- Leaderboard for each storm
- "Recent storms" leaderboard
- All-time leaderboard
- Nicer UI

Submit forecasts at: https://airtable.com/apprKdSgmNDRnsrZ9/ ... GSwK9/form (Airtable signin needed)
WORK IN PROGRESS - View forecast list at: https://airtable.com/apprKdSgmNDRnsrZ9/ ... xh1GT1HcFH
NOT DONE YET - view leaderboard at: https://airtable.com/apprKdSgmNDRnsrZ9/ ... wBcVJZ1Eev - this has some test data in it so you can see how it'll work

-------

The way scoring works (or will work once it's done) is:
- Each forecast point (12, 24, 36, 48, ...120 hours) is scored for both wind and track on a logistic curve based on the historical NHC track error at that lead time. (This is similar to, say, how Geoguessr scores you based on how close you get a location right).
- Exactly the historical track error gets 50 points.
- Less than historical error gets more, up to ~98 points for no error.
- More than historical error gets less, down to ~2 points for 2x or more historical error.
- Each point's score is the average of its wind and track scores.
- Each forecast's score is the average of the scores for its points and one imaginary zero-score point. This is to encourage not leaving harder point forecasts out.
- Each storm's score is the average of the scores for its tracks, with one imaginary zero-score forecast for every 8 best track points. This is to encourage not avoiding harder forecasts.