This paper deals with the problem of determining optimal reservoir daily operating policy over a one-year period. This problem is stochastic since the daily reservoir inflows are random and cannot be predicted far in advance. The aim of the paper is to show that optimal reservoir operating policy changes with the way the problem is solved and the information that is taken into account. The paper first shows that the operating policy determined using Stochastic Dynamic Programming greatly improves when the multi-lag autocorrelation of the inflows is included. Next, it shows that the operating policy improves with the number of days that the inflows are assumed to be known in advance. Finally, the paper shows that a better operating policy may be obtained by solving the optimization problem with Sampling Dynamic Programming. Numerical results are presented, compared and analyzed.