In this paper, the ELO scoring method is applied to rank the badminton players. By analyzing the ranking results, we find that the basic ELO model can only see the rank of the players at the top level, while other players are unable due to the reason that they have the same ELO score. Because the basic ELO model only considers the result of the competition, without taking into account the past performance of players, physical condition , and other factors, it cannot fully reflect the true level. Therefore, improving the basic model is necessary. Firstly, the initial score is rese-lected based on the past performance of players. Then, different weights are assigned to each stage of the competition, and a streak mechanism is introduced to establish a more comprehensive dynamic system. In particular , for players who are unable to participate in the competition due to unexpected factors, we also establish different scoring methods. Finally , a new badminton player ranking is obtained based on the improved model. In the final, the classical Goodman−Kruskal correlation coefficient in statistics is selected for numerical analysis, and the comparison results are highly consistent. In particular, we can measure the winning rate of different players through the model, which can also help us better predict the performance of players in the game.