A combination of many models is most useful

The Scenario Modeling Hub ran 16 different rounds of COVID-19 models over two years of the pandemic, each trying to project what the future would hold. Different models, named A through J, performed better or worse over the various rounds. But the ensemble built from multiple models was more accurate and more reliable over time.

A rating of 1 means the model performed “on average” compared to all the other models, with a lower number being better than average and a higher number being worse.
Chart: The Conversation, CC-BY-NDSource: Howerton et al, Nature Communications 2023Get the dataDownload imageCreated with Datawrapper