In this experiment, we use a genetic algorithm to co-evolve populations of simulated pursuer and evader robots for the game of “freeze tag”. The experimentation involves a virtual environment to see how our simulants perform. The first and foremost task is to develop the genetic set of all possible behaviors and responses that the robots will possess, and also to develop a sufficient evaluation function to test how well each individual behavior set performs. When this is accomplished, we take a random sampling of genotypes, and from this population, evolve increasingly better performing behaviors for the robotic evaders and pursuers. We then experimentally analyze whether the robot teams perform better with homogeneous or heterogeneous genotypes and examine the cooperation level among the players. While the robots can be steered towards cooperation, depending on the evaluation function used, we aim to see if some sort of self-organization develops in the system without specifically directing the evolution towards such.