By adopting the transition probability matrix of Markov chain method, this paper attempts to predict the short-term wave height information. This study presents the discussion on the error of wave forecast by Markov chain theory, so as to understand the capability of this theory for wave forecast. The wave forecast results are obtained from wave height transition probability matrix; it shows the mean errors of the prediction of wave height within 3 days forecast are all less than 30 cm respectively. For the sake of the decrease of the error of wave forecast, the joint probability of transition matrix of Markov chain which based on the observed wind speed and wave height data is applied in this study. It shows the method of joint probability of transition matrix of Markov chain is suitable for wave forecast in the Beaufort scale 1-6 grades. Various wave forecasts result in different seasons are also discussed here; the accuracies of the forecast wave height are relatively high in winter. There might be a great error of wave forecast during typhoon duration due to its non-stationary characteristics, which influences the accuracy of transition probability matrix of Markov chain.