Update Apr 7, 2020 2:10 PM
It’s now turned into two figures. The first one, which consists of four subplots, charts the cumulative data for deaths, positive test results (positive for the disease, not the person), completed tests, and the ratio of positive tests to completed tests. The second graph does the same thing, but for each day’s values. See this post for why daily information can be more useful.

Because the data no longer plot as straight lines, I’ve changed the light blue lines to a robust local regression fit (lowess). The purpose of the lines is to guide your eye; they aren’t intended to predict the future.

Update Mar 25, 2020 5:55 PM 
Well, as you can see, I couldn’t stop fiddling with the graph. It now includes three subplots, the original cumulative case count (positive tests for the virus) at the top, the count of completed tests for the virus in the middle, and the ratio of positive tests to completed tests at the bottom. The top two are plotted with vertical log scales, the bottom has a vertical linear scale. The horizontal scale is the same for all of them, so to save space the dates are shown only at the bottom.

I do have a mortality plot, too. I just don’t know when or how I’ll post it.

Update Mar 29, 2020 7:58 PM
Yes, I’ve changed it again. Some of the original description is now flatly wrong, and I think I’ll be changing it further, as I’m not happy with some of the layout. At some point I’ll have to change to base-10 logs, both to handle the scale and to make the tick labels easier to read.

Update Feb 2, 2021 8:26 PM
I’ve stopped updating this, which I should have done a long time ago. Neither log scales nor cumulative graphs have been useful for months.