In linear air gap analysis of semi-submersibles, the surface elevation is modified by an asymmetry factor to account for the crest-to-trough asymmetry related to wave nonlinearity and the effect of nonlinear diffraction. The asymmetry factor varies with numerous conditions including the position relative to the semi-submersible, the sea state, the relative wave direction and the semi-submersible’s loading condition. Although simplified values for the asymmetry factor are suggested in rules and guidelines such as DNVGL-OTG-13, model tests should ideally be performed in order to get accurate values for each specific design. When considering nonlinear responses, model tests are generally done according to the contour line approach, where several realizations (seeds) of a few critical sea states are performed using the 3-hour maximum from each of these to fit a Gumbel distribution. We here seek to explore if the number of seeds can be optimized when considering the relative wave elevation (upwell) by using the less data-wasteful Peaks-Over-Threshold method.