There is a persistent performance gap between the various dev* forecasters and the "smoothed scaling" forecasters (pseudopolitical.byzantium, polar.ergophobic, and pleasurable.nonmedical below). This is despite nominally using the same forecasting methods.
@brookslogan did the setup work to get single forecast comparisons between the two; the final versions of these scripts will either be included here or in covid-hosp-forecast.
There is a persistent performance gap between the various dev* forecasters and the "smoothed scaling" forecasters (pseudopolitical.byzantium, polar.ergophobic, and pleasurable.nonmedical below). This is despite nominally using the same forecasting methods.
@brookslogan did the setup work to get single forecast comparisons between the two; the final versions of these scripts will either be included here or in covid-hosp-forecast.