COVID models still don’t work

David Wallace-Wells writes,

Looking back, you could find a few lonely voices suggesting winter would be calmer than autumn. But the CDC aggregates and showcases 26 pedigreed models predicting the near-term course of the disease. On January 18, only two of the 26 showed the dramatic case decline the country experienced by February 1 as being within what’s called the 95 percent confidence interval. In other words, 24 of the 26 models said what ended up happening over just the next two weeks was, more or less, statistically impossible. The other two gave it, at best, a sliver of a chance.

Pointer from Tyler Cowen.

The article correctly points out that the only clear differences in outcomes are between Asia and the West. I still wonder whether the two faced the same virus.