The influence of ensemble size and composition on the performance of combined real-time COVID-19 forecasts.

Including more models in European COVID-19 Forecast Hub real-time ensemble forecasts improved and stabilised aggregate performance, while selecting better component models and model diversity showed no clear performance advantage.

• Why it matters: Organisers of multi-model platforms lack practical guidance on how ensemble size and composition affect forecast accuracy and consistency, especially when contributions are voluntary.
• What they did: Using European COVID-19 Forecast Hub data, they compared ensemble construction choices: larger versus smaller model pools, equal weighting versus selecting better-performing components, and diversity measured numerically or qualitatively.
• The result: Aggregate ensemble performance increased and became more stable as more models were included; however, performance did not show clear gains from performance-based selection or from greater model diversity.

AI summary of the abstract · may contain errors · not medical advice · read the paper.