Maximizing the long-term value of hydropower generation requires management of uncertain reservoir inflows, potentially variable constraints on outflows, and exposure to possibly wildly varying power prices. We describe a stochastic dynamic programming approach to the quantification of reservoir reliability (for example, measures of the risk of over-topping the reservoir or failing to satisfy downstream flow requirements) and a related approach to determining the reservoir flow strategy that maximizes expected revenue, subject to defined target reliability levels.