With the development of computer game technologies, gameplay becomes very realistic in many sports games, therefore providing appealing play experience to game players. To get the victory in a football pitch, the team composition is pretty important. There is little research on the automatic team composition in sports games particularly in a popular game of Pro Evolution Soccer (PES). In this paper, we consider the team composition as one team player recommendation problem since a team is composed of several players in a game. Subsequently, we aim to recommend a list of sufficiently good football players to game players. We convert the team player recommendation into one optimization problem and resort to the greedy algorithm-based solutions. We propose a coverage function that quantifies the degree of soccer skills to be covered by the selected players. In addition, we prove the submodularity of the coverage function and improve a greedy algorithm to solve the function optimization problem. We demonstrate the performance of our techniques in PES2018.