In different multi-agent systems, agents coordinate with different fashions. In team-based games, agents coordinate to enhance the collaborative behaviors of a team. An attractive problem in team-based games is the role assignment problem. It requires agents to decide which roles they should take based on the real-time feedback from a dynamically changing environment. The Minority Game, widely used in modeling financial marketing problems, has very similar characteristics which meet the fundamental requirements of the role assignment problem. In this paper, we propose a formulation of MG strategies in the role assignment problem in a particular multi-agent system: Simulation Robot Soccer. Through experiments, we demonstrate that MG strategies improve the effectiveness of the role assignment among agents in a real competitive environment. The improvement obeys some characteristics discovered in the theoretical MG model.