On Multi-Agent Learning in Team Sports Games