An accurate stiction model enables the detection, quantification, and compensation of this nonlinear phenomenon in a control valve. Compared with stiction models obtained from physical laws, data-driven models are more popular because of their simplicity in terms of the number of parameters required. In this work, the previously proposed two-layer binary tree data-driven stiction model is revised to overcome its limitations in handling instantaneous input commands on reverse motion. Then, the accuracy of the revised model is validated using the full set of ISA control valve standard tests. From these results, its advantages over other major existing data-driven models are indicated in terms of simplicity and accuracy.