Realtime winner prediction model for Soccer

Abstract

The objective of this study is to develop a model which can predict the in-game probabilities of winning for either team or a draw in a soccer match based on the events that have transpired at a given point of time. One of the key questions in soccer has been accurate prediction of match outcomes based on the information at hand. While traditional forecasting techniques have been limited to one time pre-match prediction, the advancement of computational prowess and forecasting techniques has led to efforts in the direction of real time predictive analysis. For this purpose, in addition to external parameters such as team form, location of match, player injuries etc., the in-game match parameters such as number of goals scored, corners, cards etc. are also of prime importance. The objective of our research is to come up with a novel, reliable and accurate real-time prediction model which can be used to predict the outcome of soccer matches, basis commonly observed match parameters.