The main aim of this research was to characterise the dynamic impact behaviour of the sliotar core. Viscoelastic characterisation of the balls was conducted for a range of impact speeds. Modern polymer balls exhibited strain and strain-rate sensitivity while traditional multi-compositional balls exhibited strain dependency. The non-linear viscoelastic response was defined by two values of stiffness, initial and bulk stiffness. Traditional balls were up to 2.5 times stiffer than the modern types, with this magnitude being rate-dependent. The greater rate of increase of traditional ball stiffness produced a more non-linear COR velocity-dependence compared to modern balls. The dynamic stiffness results demonstrated limited applicability of quasi-static testing and springtheory equations. Analysis of ball deformation behaviour demonstrated that centre-of mass displacement and diameter compression values were not consistently equivalent for all ball types. The contribution of manufacturing conditions to ball performance was investigated by conducting extensive prototyping experiments. Manufacturing parameters of temperature, pressure and material composition were varied to produce a range of balls. Polymer hardness affected stiffness but not energy dissipation, with increased hardness increasing ball stiffness. The nucleating additive influenced ball COR, with increased additive tending to reduce ball COR, but this effect was sensitive to polymer grade. The impact response of the ball was simulated using three mathematical models. The first model was shown to replicate ball behaviour to only a limited degree, despite being used previously with reported success for other ball types. The second model exhibited a reasonable representation of ball impact response that was universally applicable to all tested ball types; however, the accuracy in terms of predicting force-displacement response was not as high as required for broad range implementation. The third model exhibited significantly better accuracy in simulating ball response. The force values generated from this model exhibited up to 95% agreement with experimental data.