I see that yesterday the projected total death count was raised to almost 69,000 in the IHME model. Last time before that I looked at it it was under 62,000. Who knows where we’ll actually end up.
I imagine their model is somewhat similar to technical analysis of a stock. They are refining their curve every time that they get new, hard data.
The big difference would be that they are projecting their curve into the future based on an iffy R-naught number, that being the basic reproduction number of the virus; how many people each carrier can be expected to infect. That R-naught number is guesswork on a virus with no previous record to build on.
Here’s a good short video giving an idea of the difference that various R-naught numbers make: