In this paper, we build a model to make a reasonable prediction of some data of the game, which can help the game company to make constant adjustments to the game in its operation. For the prediction of the number of participants, we use function fitting, gray prediction (Markov chain correction) and BP neural network to predict the number of participants in the game at a specific time. The model built by the gray prediction (Markov chain correction) and BP neural network fits better. Then we introduced three difficulty influence factors for word difficulty and established a set of well-fitted evaluation criteria by linking the four influence factors with two parameters of the weibull function. Finally, we used hierarchical analysis to assign values to the difficulty influence factors, and then evaluated the words according to this assignment result