Shotifier: A Binary Shot Conversion Classifier Pipeline for Football Forwards

In this paper, we present Shotifier, a binary classification pipeline based on the principle of hybrid parallelism. Shotifier pipeline focuses on forwards or strikers, and uses the match statistics, spatial factors like start and end position of the ball, and tactical factors like events, subevents for classifying whether the shot converts into a goal or not (shot conversion). As football is a low scoring game, the emphasis is scoring the winning goal, and due to the unpredictable nature of this game, there are various factors that the clubs' coach must consider, such as team-formation, weather conditions, lineup, previous results, and many more. However, identifying the critical factors that have a substantial influence on the player's position for the shot conversion is a challenging task. Hence, in this paper, we focus our research on forwards or strikers to predict the influential factors on shot conversion. Shotifier pipeline is a hybrid parallelism based pipeline that follows a two-step approach: First, applying Kernel Density Estimator to visually represent high concentrated zones on the football pitch for specific events in the match, along with identification of maximum ball activities based on observed ball's start and end position. Second, using the historical match statistical data to identify the factors that have a strong influence on shot conversion by applying Support Vector Machine, XGBoost, Random Forest, and Multi-layer perceptron models. Finally, we use k-fold cross-validation to evaluate the effectiveness of the models and observe that the precision of XGBoost classifier is 70.39%, which is better than the other models considered in this study.