A Method for Generating Basketball Defensive Runs by Combining VAE-WGAN

The Report on the Development of Basketball in China states that the proportion of the population aged 6 to 65 who play basketball is 10.9% and the proportion of those who play basketball as their main sport is 6.7%. It leads all ball sports and is at the forefront of all sports. The right basketball tactics are decisive for the direction of a basketball game. Traditional basketball defensive tactics are usually designed based on the experience and intuition of coaches, lacking objective data support, and are often not as effective as they could be. In order to provide more scientific and effective court runs for defensive players, this paper proposes a method to generate basketball defensive runs using the VAE-WGAN network model. VAE optimises the reconstruction loss and the potential spatial regularisation loss, allowing the generated trajectories to be diverse while maintaining realism. We trained and tested the proposed VAE-WGAN model and two other tactic generation models on the same self-constructed dataset. Experiments showed that the generated defensive runs fit the real data better, and the predictive ability of this model improved by 3.7% compared to the imitation learning model proposed by Thomas Seidl, validating the feasibility of the algorithm. In addition, this paper designs a UI interface to visualise the tactical trajectory. The coach can design offensive plays on the electronic tactical board: draw the dribbling route, passing route and running route of the offensive player, click to generate them, and the system can predict the reaction of the defensive player accordingly and generate the corresponding defensive running position. The method proposed in this study can be used to simulate offensive and defensive scenarios on the court and explore the running positions of defensive players in various offensive scenarios.