In addition to point forecasts, the participants were encouraged to supply the prediction intervals of their forecasts, so that the expression of uncertainty was also captured.

These could be computed either with a sound statistical basis (e.g., when using statistical methods) or empirically (e.g., when using non-statistical or machine learning methods).

We note that the submission of the prediction intervals was in order not to discourage any participant for participating in case of not being able to provide them. Thus, the performance of the intervals submitted was evaluated separately for academic/scientific purposes.

More details can be found in the M4 – Competitors Guide.