This paper presents a study on the best possible use of optimization models for the short-term hydropower scheduling problem. Different deterministic and stochastic models have been proposed to schedule turbines for an entire planning horizon. In this paper, we take a different perspective on the problem by investigating how to determine the best optimization model during the planning horizon in order to maximize the total energy production. Specifically, we use a mixed-integer linear programming formulation as a deterministic model, where the inflows are represented by a median scenario for each day, and a stochastic multistage mixed-integer linear programming formulation as a stochastic model, where the uncertainty of the inflows is represented using a scenario tree. For the stochastic model, the stages are aggregated to reduce the computation time. We formulate the problem of choosing the best optimization model as a blackbox optimization problem. Computational experiments with two powerhouses in series, each with five turbines, for a 10-day rolling-horizon are presented. The solutions provided by the blackbox solver are analyzed, including an evaluation of the expected inflow volumes. These analyses shed light on the selection of the most appropriate optimization model. The results show that the choice of optimization model can indeed be influenced by the observed variability of inflows.