Predictive p-persistent CSMA protocol is an important MAC solution developed for networked sensor/control applications and used in Local Operating Networks (LonWorks) technology. The protocol uses a built-in network load prediction to support collision avoidance. The paper presents an analytical study of prediction efficiency for a channel with collision detection. The approach based on Markov chains is applied. The procedure of performance analysis includes the definition of transition probabilities of Markov chain for a specified load scenario, calculation of stationary distribution of contention window, and the probabilities of successful/unsuccessful transmission. It is shown that the predictive p-CSMA protocol manages to control the size of a competition window in order to guarantee the sustained probability of a successful transmission. The simulative validation of analytical results is provided.