Fig. 2
From: Using prediction polling to harness collective intelligence for disease forecasting
Example of an individual forecasting problem (IFP) of type “range”. This IFP was open to forecasting for 21 days. It featured 4 outcomes at the start, but one was ruled out by ground reports on day 12 and another on day 16. The top chart shows how the crowd-forecasted probabilities for each outcome evolved over time, with the solid line indicating the correct outcome (based on a ground truth of 48 WHO member states). The bottom chart shows the daily Brier scores (forecasting errors) of the crowd’s forecast as compared to the “chance” forecast which assigned equal probabilities to all outcomes not yet ruled out