This paper uses a new optimization algorithm employing a neural network control system that swarms with game decision-making (NNPS-C) to solve global optimization problems. The transfer function is added to the original feedback loop, it is combined with the PID controller to form a single-neuron controller, and multiple single neurons are connected to form a neural network controller. Concurrently, the fill function allows the control system to jump out of the local optimum. The grouping method is similar to the company’s organizational structure; it divides neural network subsystems into several groups with a pyramidal structure. In the meantime, the subsystems of each layer are connected with other subsystems via game decision-making, including the connection with the farthest layer. Finally, five test functions are used to prove the effectiveness of the proposed algorithm.