In conventional practice extreme design conditions for offshore structures are obtained very conservatively by extrapolating 3-hourly statistics of wind, wave and current data to a (say) 100 year return level, assuming that the 100-year extremes occur simultaneously and act in the same direction. This study involves an alternative approach accounting for the joint probability and directionality of wind, waves and currents. Design conditions are generated from the statistics of extreme global loads in individual storms, resulting in a 100-year base shear and overturning moment. Treatment in terms of storms avoids the dificulties arising from correlation between successive 3-hour intervals. The base shear forces are determined by a loading model, an analytical relationship between base shear and crest height, and most of the important environmental parameters. The inverse of the crest height-base shear relation is used to derive from the crest elevation statistics a cumulative distribution of the extreme base shear for individual storms. This is done for every storm from the northwest quadrant in the 25 years of hindcast data base in the North European Storm Study (NESS) for one location in the northern North Sea. Each storm is characterized by its most probable extreme base shear value, Fmp . These representative storm parameters are used to describe the short term and the, long term statistics of extreme base shear. It has been found that the short term variability of all storms can be well represented by one model distribution, p(f l Fmp ). With this probability distribution for the model storm, in combination with the results of a new asymptotic method estimating the probability distribution of Fmp , P(Fmp ), the probability distribution of the largest base shear for any random storm, p(f l any storm), is determined. The same analysis is followed for overturning moment. Since the average arrival rate of the storms is known the probability distribution of the largest base shear (and overturning moment) with a return period of 100 years, f100 (and m100 ), can be deduced. From a back calculation it appears that the resulting design loads are caused by combinations of extreme wind, waves and currents, which are significantly less severe than the values conventionally used for design.