Lots of people here are pointing to Philip Tetlock's book Superforecasting which describes these forecasting problems, and also a more rigorous, scientific approach to forecasting.

Interestingly, Tetlock's 'superforecasters' predict a 30% chance of more than 350,000 coronavirus deaths in the US: https://goodjudgment.io/covid/dashboard/

This is up from 17% on March 21