Competition Results Prediction Model Based on Athlete Ability Data Simulation and Analysis

This paper offers formalized description of outcome prediction for sports competitions. It proposes a novel match prediction model and team model based study on existing technologies. The team model adopts improved Bayesian network. It adopt EM algorithm to learn player capability of each team to evaluate the ability value of player, according to integrate match situation. Then the scoring probability and team score expectation are computed, combined with the case of players in presence. At last the score estimation of round prediction by match model is acquired. During study of team model coefficients, we assume it obeys distribution of Logit and Probit and compare their effects. Experimental analysis adopts three seasons of real dataset in CBA league, and our model is compared to Bayesian network model. The results show our model is more accurate than existing model in match process prediction.