There is a lot going on here and lots to criticize in the forecasts, but overall prediction markets could be really promising for disease forecasting

Here's the paper if you prefer to read that rather than my summer in the blog post above: www.pnas.org/doi/10.1073/...

In the paper we show how a single parameter addition with some epidemiological peak dynamic modeling seems to be able to help improve forecasts from every model we tested on (including the COVID-19 ensemble model)! We were really surprised that the approach rarely led to worse forecast performance.

This is actually fairly normal for measles, with countries often coming in and out of elimination status (figure from the paper showing those dynamics). It's important now to redouble vaccination efforts to get elimination status back on track!

John talks about the route of exposure with ingestion leading to high viral counts, and an inverse of this has been the possible low viral load when people have gotten eye-based infections. I think that could also contribute to the lower mortality seen in the US

An example from our now-defunct COVID-19 forecast dashboard: covid-19.tacc.utexas.edu/dashboards/t...