An important class of decision problems involve the selection of a policy in the presence of both uncertainty regarding future eventualities and a system exhibiting complex policy response. Within this paper, we examine the performance benefits of performing strategy selection for such problems using a hybrid modeling approach that combines decision analysis and system dynamics. Within this approach, a modeler uses a decision tree to encapsulate choices, uncertainties, and consequences (the last computed by a system dynamics model). While this hybrid technique offers many additional advantages in terms of expressiveness and the encouragment of systematic investigation of policy space, this paper focuses on the performance gains it provides. In particular, the use of decision trees to represent such decision problems permits the use of dynamic programming, which can dramatically decrease the costs of identifying a preferred policy beyond what is possible by evaluating possible policies in turn. The paper quantifies these performance advantages by means of recurrence relations for arbitrary trees, and derives inductively proven closed-form expressions of performance gain for complete trees. The results suggest that the hybrid method yields speedups exponential in the depth of the tree for both complete and randomly generated trees.