My tracking of actual daily new COVID-19 cases in U.S. and projections.
3/1 = 7
3/5 = 63
3/10 = 290
3/15 = 737
3/20 = 5,594
3/25 = 13,335
3/30 = 20,353 (25,000 median predicted)
3/31 = 24,742 (24,350 median)
4/1 = 26,473 (26,550 median)
Coronavirus Update (Live): 966,939 Cases and 49,295 Deaths from COVID-19 Virus Outbreak - Worldometer
Median Predicted New Case Numbers Next 5 Days
4/2 30,334
4/3 33,030
4/4 37,310
4/5 41,860 - slight increase from 40,430 yesterday
4/6 45,420
Why I'm tracking this: if we get 1,000 more cases today than yesterday, no one is going to come on TV and tell you that is good news statistically. They will say that is a big number and worse than yesterday. However, that would be a good sign that the transmission rate continues to fall.
Even if the numbers increase 1,000 a day and we are at 31,000 cases on 4/6, the reporting won't change, but that will mean transmission rates are dropping and every fraction knocked off that rate buys time and saves human lives.
My model is very close to the actual next day numbers now so I won't generate ranges. This is for several reasons:
1) I noticed a pattern in the way tests get reported, which makes the three days in the middle of the week have larger numbers and the other days smaller.
2) The model has a lot of data to draw on.
3) The transmission rate is still falling, but the fall in rate is steady not accelerating as fast as it was earlier.
4) All models get to high margins of error for any distant prediction, so the variance from the number 5 days out tells more that the variance from tomorrow's number. That is why, if you think your Governor's closing things in May and June isn't data driven decision making, you are correct. That is driven by politics and logistics. It may prove to be the correct decision, but is has no more certainty than a bet we might make on the November election.
Testing data resource:
The COVID Tracking Project
212k positive tests out of 1.17 million tested.