Cases are a really bad measure. We don't do randomly sampling in each state and we don't have any antibody tests (to see people who might be exposed, but have no longer have viral load). I found the best measures to be fatalities and the exponential scale graph by Aatish Bhatia (which uses case data, but with a lot of caveats).
> And two of the hardest-hit areas in the nation — New York City and Los Angeles County — released guidance earlier this week encouraging doctors not to test patients unless they think the test will significantly change their course of treatment. That means that potentially more people in both places could be admitted to hospitals with severe respiratory symptoms and recover — or die — and not be registered as a coronavirus case.
https://covid19data.ml/