Soccer robots have been frequently used to validate models of multi-agent systems, involving collaboration among the agents. For this purpose, many researchers in robotics have been developing robotic soccer teams which compete in events such as RoboCup. This paper presents a strategy model for multi-robot coordination in robotic soccer teams involving ball position, team member position and opponent position for the selection of a team tactic and the player roles. This assignation is dynamical and achieved by a virtual coach. This strategy model was validated in a RoboCup Small Size League environment using Webots robot simulator.