The interaction between attacker and environment predicts successfulness in one-on-one dribbles in male elite football

The analysis of individual actions in football (soccer) focuses on shots and passes. This disregards dribbling as a third option in ball possession. Successful one-on-one dribbles potentially lead to outplaying defenders and disrupting the opponent's defensive structure, which relate to winning matches. This study aims to find determinants of successful one-on-one dribbles and to predict their success using machine learning. One-on-one dribbles, pre-collected from event data, were manually labelled on their validity and successfulness leading to a sample of 734 one-on-one dribbles. Using tracking data, 24 individual, interaction, and environmental variables were obtained and used to predict and differentiate the outcomes of one-on-one dribbles. Variables related to player interaction, like the attacker-defender distance (p < 0.001, r = 0.27), the attacker-ball distance (p < 0.001, r = 0.36), and the pressure on the attacker (p < 0.05, r = 0.24), differentiated between successful and unsuccessful one-on-one dribbles in the multivariate analysis, while individual and environmental variables did not (p > 0.05). A light gradient boosting classifier performed best in predicting the success of one-on-one dribbles (AUC = 0.69). Our results show that success in dribbling actions can be predicted, similar to shooting and passing actions. Furthermore, the interaction between the attacker and its environment is most decisive for successful one-on-one dribbles.