A Model for Improving Guesses of Future Soccer League Game Results

Abstract

In this paper we propose a model for predicting future soccer game results by using information about the results of past league games. First, from regressions we extracted the prediction confidence interval for the goal difference between the winner and the loser in each game. Second, we created an arbitrary range around zero and defined criteria for forecasting a win, a tie or a loss according to the location of the confidence interval relative to the arbitrary range we defined. Third, we gradually changed the edges of the arbitrary range and repeated the second step. Among all the arbitrary ranges, we chose the one that best predicted the match results. We found that the best arbitrary range accurately predicts 52% of the match results. Finally, we upgraded the model by allowing double chance betting, which offers gamblers five possible betting options: home team wins (1), home team wins or game ends in a tie (1 and X), away team wins (2), away team wins or game ends in a tie (2 and X), game ends in a tie (X).  When double chance betting was allowed, the model accurately predicted 77% of the match results.