Forecasting the Next Shot Location in Tennis Using Fine-Grained Spatiotemporal Tracking Data

Memory Augmented Deep Generative models for Forecasting the Next Shot Location in Tennis

“…They utilise handcrafted, dominance features together with the ball bounce location, ball speed and player feet locations when determining the future player behaviour. This model is further augmented in [5] where the authors utilise a Dynamic Bayesian Network to model the same set of features.…”

“…The system records (x, y, z) positions of the ball as a function of time, along with the player feet positions at millisecond granularity, and other meta data including current points, time duration, sever and receiver. The dataset consists of around 10,000 shots, however as the tournament progresses in a knock-out format, similar to [5] we focus our analysis on the top 3 players.…”

“…We would like to compare it against the model of Wei et. al [5], where the former model has the attained the classification process through an automatic feature learning method. Comparatively similar accuracies emphasises the importance of the hierarchical feature learning process of deep learning models, which automatically learn semantic correspondences through back propagation.…”

“…The model in [45] improves upon this via hierarchical feature learning from the ball trajectory characteristics through a recurrent model. The method of [5] builds upon the trajectory features, via incorporating the game and player context into the prediction framework, however, fails to capture salient information from longterm dependencies among player behavioural patterns. In contrast, we capture those hierarchically, allowing us to effectively propagate this information into the future action generation pipeline.…”

“…Several studies have shown that this reactive ability is the product of pattern recognition skills that are obtained through a "biological probabilistic engine", that derives theories regarding opponents intentions with the partial information available [1], [5], [6]. For instance, it has been shown that expert tennis players are better at detecting events in advance [1], [7] and posses better knowledge/ expertise of situational probabilities [3].…”