Proper prediction of long-term extreme values for operating wind turbine loads and deflections is a critical component of wind turbine design. Direct observations or simulations of long-term extremes are not yet available; therefore, these predictions rely on some combination of large numbers of simulations and extrapolation. Extrapolation methods themselves can have significant uncertainty, and they also require that the wind turbine designer have a greater level of statistical expertise—factors that make the methods less attractive for industrial application. As an alternative to extrapolation, safety factors can be calibrated using techniques that allow designers to use smaller data sets. To calculate such factors, a series of simulations was used to extrapolate 50 year extreme values for a 5 MW wind turbine. Two methods are proposed for calculating such safety factors: one based on the mean and standard deviation of extreme values, and one based on the median of extreme values. Through a process of random sampling without replacement, the safety factor based on the median of extreme values was found to be less variable and also more independent of the number of simulations. The safety factors required were as large as 1.7, or were only 1.25 if rotor thrust loads were considered the dominant design drivers. Copyright © 2008 John Wiley & Sons, Ltd.