Number of deaths has been outside the IHME prediction intervals 70% of the time

I had to read Table 1 like 10 times to understand what it says. It's an horrible way to present the data.

The Figure 2 is strange. It's a graph of the error in the forecast for 1 day and the error in the forecast for 2 days. It´s not surprising that is is correlated. I´m not sure if it's clear why it is so linear, but perhaps it is necesary to read the underlaying model to undestand that. Anyway, if it is so linear, perhaps it may be useful to use as a empiric correction.

I should undust my probability books, but IIRC the number of cases should follow a Poisson distribution, with a different lambda each day. In that case the 95% interval should be huge, like from almost 0 to 3lambda. It looks difficut to miss an interval that is so huge.