The early projections accepted "confirmed cases" as their baseline for total infections in calculating total hospitalizations and mortality, when every person I know who works with big data understood immediately that those numbers were unreliably low. The confirmed cases number suffered from severe selection bias and limited testing availability in every case (although the limits and criteria for testing varied from place to place as well as over time). The models that assumed that 70% of the world would get this bug should have had the greatest upward adjustments to estimate actual cases based on confirmed cases because the more contagious the bug the less likely confirmed cases is capturing the entire population. A first year PhD student in any subject should be able to identify this potential issue and those early projections would have been laughed out of the peer review process.

The approach that was adopted resulted in crazy numbers (i.e. deaths in the millions in the US). Additionally, a number of the projections were "leaked" with only partial information that was primarily the worst case scenarios. These scenarios were released without the median (most likely outcome), the confidence intervals, or any other information that would be helpful with interpreting the data released. It appeared to me that it was calculated to illicit the greatest social response under the guise that the policy end justified playing loose with the data. Even as they were saying it they already made the excuse that if people change behavior then their projections will be wrong, but they'd have been way off even if we had all continued with business as usual. The projections are getting better with the suggestion that distancing efforts are bringing the numbers way down, but these numbers are are a lot closer to what they would have been from the beginning if they had employed proper techniques up front than to the initial projections. Maybe they saved a lot of lives by going for the shock factor, but I think time will tell if the remedy is actually any better than the disease and I don't see any scenario where history looks favorably on the early projections that never had a chance of being correct.