Tactical behavior of soccer players always comes with both risk and reward depending on the outcome of the actions. This is especially true during the often chaotic playing phase of transition following a ball gain/loss, where player actions can highly impact a team's success. Therefore, this study analyses the risks and rewards of transitions, focusing on the objectives of both the defensive and offensive teams. Official tracking and event data of 612 matches played during two Bundesliga seasons (2022/23 and 2023/24) were analyzed. To assess transitions, two separate machine learning models were developed based on over 30 expert-driven features: (i) an expected possession value (EPV) model that predicts the probability of the ball-gaining team to score a goal in the following seconds of the possession, (ii) an expected ball gain (xBG) model that predicts the probability of the ball losing team to regain the ball in the following seconds of the possession. Overall, 220,226 match situations during the 58,868 transitions were analyzed. The resulting EPV (AUC: 0.88) and xBG (AUC: 0.64) models showed satisfactory prediction performance. Based on the combination of both models, the risk and reward of match situations for both teams can be quantified in detail which ultimately assists in analyzing the tactical strategy in offensive and defensive transition. This is illustrated in an in-depth individual player analysis to evaluate tactical player decision-making and determine the influence of the players on the success of a team.
Forcher et al. (Fri,) studied this question.