The importance of recent scores in a forecasting model for professional golf tournaments

Abstract

In many fields of forecasting, practitioners face the problem of how to distinguish signal from noise in recent news. As an illustration of how one might approach the problem we examine, in the context of golf, the extent to which recent players' scores embody genuinely new information relevant to predicting the outcomes of tournaments. We construct an ordered logit forecasting model for tournaments on the Professional Golf Association US Tour. Player scores in their six most recent appearances are demonstrated to have an important role additional to that of a longer-run indicator of player quality. The model performs well relative to reliance on weekly official world rankings of players and the results of log-likelihood tests indicate that this is because of greater weight attached to very recent tournament performances.