In many offshore operations, including the operation of wave energy converters, phase-resolved wave prediction is essential for safety and efficiency. However, for sea-states with large directional spreading angles, wave prediction becomes increasingly difficult. This study investigates the benefit of exploiting concurrent surface-displacement time histories measured by a buoy in three degrees of freedom (DoF) to facilitate prediction in such highly spread sea-states. A direct comparison is made with arrays of sensors measuring in a single DoF using the same formulation. The arrays considered are extensively optimised along with multiple sets of representative angles used to describe wave propagation, which is assumed to be normally distributed. The performance of the scheme is tested on a range of synthetically generated, realistic sea-states comprising single and crossing swells prevalent along the south coast of Australia. Additional tests involve wind-sea and swell waves described using Ewans models. Accurate predictions are obtained for times up to three wave periods into the future, in typical cases. With the multiple sets of optimal angles obtained for each optimal array, we further enhance the capability of the scheme by aggregating and averaging the different predictions from these sets. We observe a marked increase in both performance with large spreading and tolerance to possible variation in mean wave directions.