This work introduces an efficient finite-position-set model-reference adaptive system (FPS-MRAS) estimator for sensor-less control of a doubly fed induction generator (DFIG) in a variable speed wind energy conversion system (WECS). The idea of the proposed FPS-MRAS is derived from the finite-control-set model predictive control (FCS-MPC) principle, i.e. a limited set of angles for the rotor position are employed to predict the rotor current of the DFIG. Subsequently, the optimal angle for the rotor position is the one that its prediction optimizes a pre-defined (novel) quality function. Accordingly, the constant-gain proportional-integral (CG-PI) regulator and its tedious tuning which is normally used in MRAS observers is not required. The estimation response of the suggested FPS-MRAS observer is investigated by experiments and compared with the performance of the classical MRAS under different operation conditions and at variations of the DFIG parameters.