It frankly is stunning how bad they are. All I can think is people just shove some numbers in a computer and trust that what comes out the other side is right. Pure statistical models are killing these things in accuracy. Look at the most recent University of Washington model that is way more conservative than the ones before and it comes AFTER the lockdown so it has that built in - https://covid19.healthdata.org/projections

They projected today in NYC for them to have 50k hospitalizations. They only have 12k right now. They are off by even more in other states. Again, this is a model that was done AFTER the lockdowns. If you want to see someone who is getting it right, it's this guy - http://sinaas.blogspot.com/2020/02/corona-novel-outbreak-guess.html

He modeled it after a typical outbreak curve (imagine that!). He guessed on March 20th that today Netherlands would have 1325 total deaths, actual was 1173. He guessed on March 19th that Italy would have 13,700 total deaths, and they have 13,200 (and he even guessed it downward on the 25th).

I'm sticking with my prediction that the U.S. peak deaths are in 3-5 days and it will follow the same curve as everyone else.