FOOTBALL DATA ANALYSIS: THE PREDICTIVE POWER OF EXPECTED GOALS

Major football events, such as the World Cup, have popularized the European football (i.e., soccer) beyond its traditional geographical borders, for example with a growing number of Americans who have begun to incorporate European football into their daily sport consumption. Many video game series based on this sport have further contributed to this increasing development. Not only that, but many performance indicators inspired by football computer games have become metrics that are now part of the normal conversation about real football, being used to analyse tactics, teams and players. Among these indicators, of prominent importance is the role of expected goals (xG) that measure the quality of a shot by calculating the likelihood that it will be scored from a particular position during a particular phase of play. Often, someone talks about xG leading to imprecise conclusions and to a distorded point of view on this performance indicator. In this paper, using a non-parametrical Kolmogorov Smirnov hypothesis testing method, we show that there is no statistically significant difference if, in a given football tournament, the final scores are awarded to the participating teams either with real goals or based on xGs, thus demystifying some wrong assumptions about this indicator. Data for this study come from all the matches played during the 2020 UEFA European Football Championship.