Aiming at the interaction of various contextual variables on basketball competitions, this paper puts forward the idea of analyzing the key indicators that affect the success or failure of basketball competitions based on contextual variables. K-means clustering is used to define balanced competition and unbalanced competition, and three contextual variables are included in the contest location(host/guest), type(balance/unbalance) and result(win/loss). Grey B-type correlation theory and binary logistic regression are used to establish mathematical models for technical indicators and contest results in different contest situations of CBA league. The results show that the key indicators affecting contest success or failure in different contest situations are the key indicators. There are some differences, among which backcourt rebounds and steals are the winning indicators that affect the common existence of the game. It is suggested that coaches should focus on training key technical indicators according to different competition situations.