Model analysis in system dynamics (SD) entails articulating exactly how the structure of circular, feedback relations among variables in a system determines its performance through time. This article combines human games with SD to show the use and benefits of model analysis with the pathway participation metric (PPM) implemented in the Digest®software. Four SD game models depict Markovian paradoxical games with two players or groups that choose between collegiality and discord tactics as the means of conflict resolution in business and civil litigation. Paradoxical self-referential games are non-constant sum (one player's loss is not automatically the other's payoff) conflicts, where the two players or groups compete with dynamic (i.e., time varying) probabilities of collaboration. Their game is paradoxical because both parties can either win or lose simultaneously. It becomes self-referential when the payoff or 'tempting' parameters, and the prior discord and loss coefficients depend explicitly on the participants' collaboration probabilities. Large subsets of initial discord tactics converge on a fixed-point attractor to sustain collaboration equilibria. Games end once the point attractor has absorbed all dynamics, leaving the system in a stable, negative feedback state.