Research on coke price forecasting is of theoretical and practical significance. Here, the Kalman filtering algorithm was used to analyze the price of coke. As the only state variable, the historical coke price is sorted out to build the state space model. The algorithm makes use of innovation composed of the difference between observed and predicted values, and allows us to obtain the optimal estimated value of the coke price via continuous updating and iteration of innovation. Our results show that this algorithm is effective in the field of coke price tracking and forecasting.