[furyofantares]

I agree with your general point that rate=deaths/infections is making a big assumption about unresolved cases and almost surely gives a too low number -- but rate=deaths/recovered I think does the same thing in the opposite direction. Unfortunately we just don't know right now. It would be nice to see deaths vs recovered vs mild cases vs critical cases for people who have been infected >N days, but that's much more nuanced information to consume than some "death rate".

Btw, you made a point about exponential growth, but is that happening? The logarithmic graph on https://www.worldometers.info/coronavirus/ appears flatlined to me. Maybe (probably?) that's just about containment in China, but when we're discussing death rates that's where most of the data is.