We compare the forecast performance of our SuEIR model with the forecast performance of some models developed by other universities. We calculate the prediction error of each model by subtracting the reported number (groundtruth) from the predicted number. For instance, +20 for UCLA-SuEIR means that on the specified date, the prediction of UCLA-SuEIR is larger than the reported number by 20. We rank each model based on the daily prediction error in an ascending order. Then, we use the Borda rank aggregation method to combine these daily rankings over one week to obtain the ranking of the weekly prediction performance.

More multi-model comparisons can be found at the following sites: Reich Lab, 538, CDC