Characteristic loads for design of offshore structures are defined in terms of their annual exceedance probability, q. In the Norwegian Petroleum Regulations, q = 10−2 is required for the ultimate limit state (ULS), while q = 10−4 is required for the accidental limit state (ALS). In principle, a full long-term analysis (LTA) is required in order to obtain consistent estimates. This is straightforward for linear response problems, while it is a challenge for nonlinear problems, in particular if they additionally are of an on–off nature. The latter will typically be the case for loads due to breaking wave impacts. In this paper, the challenges related to estimation of characteristic slamming loads are discussed. Measured slamming loads from a model test are presented, and the observed large variability is discussed. The stochastic nature of slamming loads is studied using a simplified linear relation between the sea states and the Gumbel distribution parameter surfaces. The characteristic slamming loads with q-annual probability of exceedance are estimated from an LTA using the short-term distribution of the slamming loads and the long-term distribution of the sea states. The effect of integrating over a smaller area of the scatter diagram of the sea states is studied. The uncertainties in response from slamming loads are compared to a more common response process, and the relation between variability and the number of realizations in each sea state is looked into.