The Stress of a Serve: Biomechanical Quantification of Shoulder Muscle Load in the Tennis Serve

Abstract

Shoulder injuries are common in tennis and are often associated with the high joint loads generated during the serve. While previous work has described kinematics, joint kinetics, and energy transfer along the kinetic chain, less is known about how individual shoulder muscles generate, transfer and absorb mechanical energy during the serve. The aim of this thesis was to investigate how mechanical energy is generated, transferred, and absorbed in the upper extremity during the tennis serve, with a specific focus on the shoulder and elbow joints and the muscle groups responsible for accelerating and decelerating the arm.
Data from five male competitive tennis players were analysed. Three-dimensional motion capture and surface EMG were combined with an upper-extremity OpenSim model that included the thorax, shoulder complex, elbow, forearm, and racket. Inverse kinematics and inverse dynamics were used to obtain joint kinematics and net joint moments. Static optimization (SO) was then applied to solve for muscle forces that generated the experimental accelerations. Muscle power and work were computed for all shoulder muscles, and generalized actuator work was quantified to assess the contribution of non-muscular actuators. The serve was divided into three time windows based on kinetic energy peaks of the thorax and forearm: (1) trophy position to thorax peak, (2) thorax peak to forearm peak, and (3) forearm peak to maximum internal rotation (MIR).
Model verification showed acceptable marker tracking accuracy for most markers, with distal marker RMSE values predominantly below 10 mm and proximal markers generally below 20 mm. Reserve actuator work remained small in most coordinates and time windows, but increased around ball impact for shoulder axial rotation and elbow flexion. EMG–SO comparisons demonstrated moderate-to-strong agreement for the pectoralis major and more variable correlations for the triceps, latissimus dorsi, deltoid, and biceps, reflecting both modelling assumptions and physiological factors.
During the early and mid-acceleration phase (trophy position—where the player tosses the ball while preparing to serve, standing as if holding a trophy—to forearm peak), the shoulder–elbow joint system showed relatively small changes in net joint work, while substantial positive and negative muscle work occurred simultaneously across muscles. The Serratus Anterior, Subscapularis, and medial deltoid consistently produced positive work, whereas the infraspinatus, teres minor, rhomboids, and parts of the trapezius absorbed energy. These patterns suggest that many muscles acted less as pure rotators and more as stabilisers within the “compressor cuff” and scapular control system, helping to manage trunk-to-arm energy transfer rather than simply increasing total system energy.
In the late phase (forearm peak to MIR), all participants showed large negative joint work, consistent with rapid arm deceleration. Muscle absorption during this phase was distributed across scapular stabilisers, abductors, horizontal abductors, and (to a lesser extent) the posterior rotator cuff. The medial deltoid and lower trapezius frequently exhibited large negative work values, indicating a prominent role in decelerating the elevated arm and stabilising the scapula. Functional grouping of muscles revealed clear inter-individual differences: some players showed a broad distribution of absorption across all groups, whereas others showed scapular-dominant or abductor-dominant strategies.
Overall, this thesis demonstrates that the tennis serve is characterised by patterns of muscle-level energy generation, transfer, and absorption that are not visible from net joint work alone. The results highlight the importance of scapular muscles working eccentrically to maintain scapular control (e.g., Rhomboids and Lower Trapezius), as well as shoulder abductors, alongside the posterior cuff, in managing deceleration loads at the shoulder. The work also illustrates how musculoskeletal modelling can be used to link kinetic-chain mechanics with individual muscle contributions, while emphasising the need for improved scapular modelling, subject-specific anatomy, and the inclusion of external forces in future studies.