This paper analyses the forecasting accuracy of experts vis-à-vis laypeople over three seasons of English Premier League matches. We find that former professional football players have superior forecasting ability when compared to laypeople. The results give partial support to the view that a crowd forecast offers the greatest precision. Pundits generate a positive return while both the crowd and laypeople generate losses. As the prediction of multiple score outcomes represents a computationally difficult task, both groups display forecasting biases including a preference toward specific score forecasts. The results are relevant for those concerned with gambling behaviour if the forecasting strategies adopted here generalise to match betting markets.