Time variation of water-retaining risk and uncertainty of filling progress aggravate the difficulties in making decision for high rock-fill dam construction in flood period. Accordingly, the water retaining risk of high rock-fill dam was calculated by stochastic simulation, and then was classified into three risk states which could indicate the decision actions. With the objective of minimizing the cost of the entire construction process in flood season, the critical factors such as incoming flood flow, filling schedule and cost of decisions were taken into account to build a decision-making model based on Markov Processes. By using this model, the flood prevention strategies, cost functions and transfer probability for each risk degree were analyzed. After testing the Markov property, the optimal strategy for each state at current time and a sequential risk decision path in this process can be predicted by strategy iteration method. A study case shows that the adjustment results by the proposed model consists well with the actual construction plan change for water retaining elevation. The results offer reference for timely and accurate flood prevention scheme in construction period.