[burger]

Interestingly, though, if you take a meta view of models the extreme outlier models are likely to have been wrong (we are unlikely to see 2 million deaths as the worst models predicted, and we already have more deaths than the most benign models predicted).  My guess is that we will probably land on the low end or slightly below the range of the original IHME model, which predicted 100 - 240,000 deaths, but probably a bit above their revised model of 60,000 deaths.

I agree with most of what you said with two exceptions:

1) I believe the Imperial College model (the one that predicted 2 million dead) assumed no change in R over time, so these disease just spread (almost) everywhere until herd immunity kicked in, and ~1% of the infected died.  That was prima facie wrong, but it also spurred quicker action by governments, so it was useful.

2) Just because the IHME model gets in the ballpark of total deaths does not make it right or even useful.  The model is predicting a very specific shape of the curve for the deaths over time, and it's been wrong in a lot of cases.

[burger]

Interestingly, though, if you take a meta view of models the extreme outlier models are likely to have been wrong (we are unlikely to see 2 million deaths as the worst models predicted, and we already have more deaths than the most benign models predicted).  My guess is that we will probably land on the low end or slightly below the range of the original IHME model, which predicted 100 - 240,000 deaths, but probably a bit above their revised model of 60,000 deaths.

I agree with most of what you said with two exceptions:

1) I believe the Imperial College model (the one that predicted 2 million dead) assumed no change in R over time, so these disease just spread (almost) everywhere until herd immunity kicked in, and ~1% of the infected died.  That was prima facie wrong, but it also spurred quicker action by governments, so it was useful.

2) Just because the IHME model gets in the ballpark of total deaths does not make it right or even useful.  The model is predicting a very specific shape of the curve for the deaths over time, and it's been wrong in a lot of cases.