In keeping with the trend towards lightweight design in the automotive industry, the demand for statistical proven material characteristics for dimensioning of highly-stressed components increases. Especially in the sectors of serial- and mass production, minor failure probabilities need to be guaranteed, whereat an accurate knowledge of the fatigue strength and the underlying scatter is required. In this thesis, a fatigue strength estimation based on the statistical size effect is presented and experimental fatigue tests are performed to validate this approach. In the first part common statistical evaluation procedures of S/N-test results were analyzed by Monte-Carlo-simulations in terms of accuracy and quality of estimation. Based on the simulative results, recommendations for the application of the investigated procedures as well as statistically-based safety factors were derived. The results reveal that for a determination in the high-cycle-fatigue regime the pearl string method should be used. If there are just a few specimen available to evaluate the long-life-fatigue strength and if only the mean of the fatigue strength is of interest, the arcsin square-root(P)-procedure performs well. In case that the underlying scatter is also of interest, the staircase-method based on the evaluation procedure of IABG should be applied. The second part deals with the execution of experimental Wöhler fatigue tests incorporating two different round specimen of variable length, consisting of a cast AlSi8Cu3 alloy. The results were assessed with regard to the statistical size effect, wherein the corresponding highly-stressed volumes of the two specimen geometries were calculated utilizing finite-element-simulations, showing a size ratio of about one to ten. Further statistical analyzes indicate a significant connection between the size of the highly-stressed volume and the corresponding fatigue strength. A final consideration of the extreme value distribution of the base material porosity forms the basis for an approach to calculate cyclic material properties for minor failure probabilities in the ppm range. This serves to a secure and economic dimensioning of cast components in the automotive industry.