Stochastic representations of turbulent wind and irregular waves are used in time domain simulations of offshore wind turbines. The variability due to finite sampling of this input loading is an important source of simulation error. For the OC4 reference jacket structure with a 5 MW wind turbine, an error of 12–34% for ultimate loads and 6–12% for fatigue loads can occur with a probability of 1%, for simulations with a total simulation length of 60 min and various load cases. In terms of fatigue life, in the worst case, the lifetime of a joint was thereby overestimated by 29%. The size of this error can be critical, i.e., ultimate or fatigue limits can be exceeded, with probability depending on the choice of number of random seeds and simulation length. The analysis is based on a large simulation study with about 30,000 time domain simulations. Probability density functions of response variables are estimated and analyzed in terms of confidence intervals; i.e., how probable it is to obtain results significantly different from the expected value when using a finite number of simulations. This simulation error can be reduced to the same extent, either using several short simulations with different stochastic representations of the wind field or one long simulation with corresponding total length of the wind field. When using several short-term simulations, it is important that ultimate and fatigue loads are calculated based on the complete, properly combined set of results, in order to prevent a systematic bias in the estimated loads. Copyright © 2014 John Wiley & Sons, Ltd.