Several formal models combine probabilistic and nondeterministic features. To allow their probabilistic simulation and statistical model checking by means of pseudo-random number sampling, all sources of nondeterminism must first be quantified. However, current tools offer limited flexibility for the user to define how the nondeterminism should be quantified. In this paper, we propose an expressive probabilistic strategy language that allows the user to define complex strategies for quantifying the nondeterminism in probabilistic rewrite theories. We have implemented PSMaude, a tool that extends Maude with a probabilistic simulator and a statistical model checker for our language. We illustrate the convenience of being able to define different probabilistic strategies on top of a system by a cloud computing example, where different load balancing policies can be specified by different probabilistic strategies. We then use PSMaude to analyze the QoS provided by different policies.