Predictive power of fantasy sports data for soccer forecasting

We analyse data from 5,000 competitors who participated in an online soccer managerial game which revolved around the English Premier League (EPL). We show that competitors incorporate into their decisions relevant information about the outcome of a soccer match. Furthermore, forecasts based on managerial game data are significantly better than random forecasts, forecasts based on relative frequency, and forecasts based on teams' attendance, but worse than bookmaker odds. Our work provides an evidence that crowds poses significant amount of information for the match outcome prediction.