Probabilistic modelling of pavement performance has become increasingly popular as it considers the stochastic nature of pavement behaviour and deterioration, as well as the imperfections and limitations of pavement condition data. Markov chains are extensively employed to simulate the probabilistic performance of pavements using numerous methodological tweaks. The existing literature lacks a synthesis of Markov chain models and their associated methodologies. A comprehensive synthesis can elucidate the differences across various Markov model forms and methods, while determining the appropriate Markov model type based on specific data types and their availability. This paper examines different Markov chain models utilised to predict pavement performance and the methodologies for estimating the transition probability matrix (TPM), an essential element of Markov models. This study provides a critical analysis of many aspects of Markov models as used in the literature, reveals knowledge gaps and offers suggestions for their resolution. This work further develops a decision tree to select the suitable Markov model type and TPM estimate method for modelling pavement deterioration based on data types and availability. This article provides guidance and decision support for researchers and highway agencies in selecting appropriate Markov chain techniques for modelling pavement performance.