Agent-based simulation is often used to analyze various socio-economic systems. Some earlier studies suggest that the dynamics may depend on the number of agents, but few studies have investigated this issue. To tackle this, we take up the minority game and carry out computational experiments to analyze whether and how increments of agents affect the system dynamics. Our results have suggested that under specic experimental setups the game efficiency depends on the number of agents. In particular, the relationship between the number of agent and specic parameters may have affected the results.