In order to investigate the influence of initial condition uncertainty on the profile of maximum wave, two typical short-term sea states are reconsidered in this paper. The associated uncertainty is investigated with the Monte Carlo Simulation method and the maximum wave is extracted from the numerical simulation performed with the fully nonlinear Chalikov-Sheinin (CS) model which has been validated before. Six parameters are defined in this work to describe the property of the profile of maximum wave and analysed with statistical method. To catch the maximum wave in each realization, it is proved that both zero-up-crossing and zero-down-crossing methods have to be utilized in the wave-by-wave analysis. After analysing the sensitivity of different variables to the initial random-phase/amplitude model, it reveals that the maximum crest, trough and crest-to-trough heights in a sea state are only sensitive to the amplitudes of a certain range of free waves (around the peak wavenumber). Briefly speaking, three types of uncertainties can be identified in the initial conditions involving the natural, knowledge and mixed uncertainties and they contribute a lot to the versatility of the profile of maximum wave. The conclusions drawn in this paper can cast some light on the research of real-time prediction of maximum wave in ocean engineering.