Today's the day when I finalize all the parameters in our model related to uncertainty and it always gets me in a mood. One thing it's important to bear in mind is probabilities get scrunched up once you get closer to 100% in ways that can conceal important practical differences.

For instance, a forecast saying there's an 85% chance of an event and a forecast saying there's 98% chance may sound similarly confident. But in the former case, the event will fail to occur around 1 in 6 times, while it's 1 in 50 times in the latter case. Big difference.

I know that's pretty basic stuff, but there's a lot of hard work that goes into distinguishing the 90 percent chances from the 95s, and the 95s from the 98s, and the 98s from the 99.5s. And you have to think carefully about the structure of your data & model for these edge cases.