This paper reviews a range of methods used to infer characteristic values and probabilities of failure from a sample of measured or simulated load effect data. The popular methods of Peaks-Over-Threshold and Generalized Extreme Value (GEV) are considered but also other methods including the Box-Cox approach, fitting to a Normal distribution and the Rice formula. These five methods are fitted to the tail of the daily maximum data. Bayesian Updating and Predictive Likelihood are also considered, but are fitted to the entire data set. In general, the five tail fitting methods are reasonably accurate at inferring characteristic annual maximum values from 1000 days of data and the other methods less so. All methods are considerably less accurate at inferring probabilities of failure than characteristic values. Review, Peaks-Over-Threshold, Extreme Value