One of the main statistics that is used to summarize the performance of a soccer team after a game is the number of shots (on target) taken by the team. A team with many shots is being seen as having exhibited a particularly offensive game plan and challenged the opponent. However, the number of shots is an aggregate metric that does not consider the quality of the shots taken from a team. For example, a set piece that resulted in a shot from the box is certainly of higher quality than a shot taken from the middle of the field. Hence, in this work we first introduce a model for calculating the probability of a shot resulting into a goal. For training our model we use a manually annotated set of shots from a subset of the 2016 regular season`s MLS games. Our evaluations show that the model is able to accurately estimate the probability of a shot resulting to a goal. Using this model we then calculate the expected number of goals for a team and consequently its offensive and defensive efficiency by comparing this expected number with the actual goals scored/allowed. Finally, borrowing ideas from fractal theory, we analyze the dimensionality of MLS teams based on the locations they take their shots from and show that teams that exhibit lower dimensionality on the field tend to have higher offensive efficiency.
© Copyright 2018 Journal of Sports Analytics. IOS Press. Kaikki oikeudet pidätetään. / Kääntänyt https://www.DeepL.com/Translator