A holy grail for military, diplomatic, and intelligence analysis is a valid set of software agent models that act as the desired ethno-political factions so that one can test the effects that may arise from alternative courses of action in different lands. This article enumerates the challenges of such a testbed and describes best-of-breed leader and follower profiling models implemented to improve the realism and validity of the agent. Realistic, `descriptive' agents are contrasted to rational actor theory in terms of the different equilibria one would expect to emerge in conflict games. These predictions are examined in two real world cases (Iraq and SE Asia) where the agent models are subjected to validity tests and a policy experiment is then run. We conclude by arguing that substantial effort on game realism, best-of-breed social science models, and agent validation efforts is essential if analytic experiments are to effectively explore conflicts and alternative ways to influence outcomes. Such efforts are likely to improve behavioral game theory as well.