In this work, the design and testing of a new LCF double notch specimen geometry is presented. The deterministic CMB model and a probabilistic model for LCF which considers size effects are simultaneously applied as shape optimization functional. In order to demonstrate the potential of the statistical size effect consideration, the probability of crack initiation is maximized in the milder notch while the deterministic LCF life is minimized in the sharp notch. This also increases the likeliness of an experimental validation success considering the issues of high scatter in material properties and limited test resources (test plan optimization). Specimens of a high-chromium forged steel for elevated application temperatures have been manufactured according to this specimen design and a systematic and advanced experimental testing campaign is conducted to prove the hypothesis. In-situ time-resolved alternating current potential drop measurements in combination with load-triggered digital image acquisition at both notches are applied in order to enable determination and quantification of crack initiation and growth at superior accuracy during LCF cycling. These experimental and further post-experimental analyses confirm that cracking in the mild notch always occurred in the first place which has been expected for most test outcomes according to the probabilistic model predictions. Crack initiation in the sharp notches on the other hand, which has been expected to occur first on the basis of conventional deterministic predictions, could not be detected at all.