Latching and de-clutching have been proposed for controlling Wave Energy Converters (WEC), for waves of period higher and lower than the natural period of the WEC, respectively. Historically, the selection of these control strategies was based on the a priori assumption of the need to match the phase of motion of the WEC collector to the incident wave force, and to minimise the applied forces. In this paper, a standard time-domain model of a WEC is controlled by time-varying damping (representing power extracted by a generator). By the application of a novel evolutionary algorithm and artificial neural networks, the time varying damping is optimised to maximise the mean power extraction in a variety of sea states, including both monochromatic and polychromatic (Bretschneider) seas. This method can be used to optimise any control strategy. The resulting time-varying damping coefficients suggest that the resultant generic control strategy in fact represents the specific cases of latching and de-clutching. Furthermore, it is suggested that both of these strategies are required in combination in all sea states for optimum power extraction. These results are achieved with no a priori assumptions regarding the nature of optimum control (with regard to forces applied of phase of motion) and thus represent an independent, implicit validation of the latching and de-clutching control strategies.