Good podcast that helps you understand what's going on with modeling. Chris Murray's IHME model is one of the pictures that was shown in a WH Press Conference.

https://fivethirtyeight.com/features/politics-podcast-how-one-modeler-is-trying-to-forecast-the-toll-of-covid-19/

Some key points for those who want the cliff notes:

• The group is part of a hospital system and the main purpose of the model is to predict peak hospital resource requirements to help them plan.
• They use death data from places that had earlier outbreaks to form curve fits for developing outbreaks
• For example, to estimate ICU/ventilator/hospital bed need, they use ratios provided by data available from deaths to ascertain that number. Confounding variables of course are different hospital policies in other countries and the issues with reporting. Early data they had did overestimate the need and the model was revised downward after they got new data from NYC.
• While the WH did use their image during a press conference, they never predicted 100,000 to 200,000 deaths. They are not sure where that number came from.
• They have been predicting 35,000 to 120,000 deaths in their probability range. They do not think people should pay much attention to daily fluctuations within that range (e.g. why 60K today vs. 68K yesterday)
• They admit there will be difficulty as some places begin to unwind social distancing measures partially.
• They are only focusing on the first wave, not trying to predict a second wave. In the past, sometimes predictions of second waves have not materialized.

Below was the model that was presented at the WH press conference Mar 23. On top of it I graphed the daily deaths line. The model has actually performed pretty well.