Long–term creep–rupture properties are usually evaluated from short–term data by time–temperature parameter (TTP) method, such as arson–miller parameter (LMP) and Orr–Sherby–Dorn (OSD) methods. However, the conventional TTP methods sometimes overestimate long–term creep rupture properties if the prediction is based on their short–term test data for 9%—12%Cr ferritic steels. The following concepts/methods are thus proposed in this paper in order to reduce the property overestimation tendency caused by the conventional TTP methods and to obtain a better agreement of the predicted property values with the observed ones. They include the C–value optimization and the multi–C region analysis, the long–term (5×10³—1×10⁵ h) creep rupture property prediction using short term test data (≤5×103 h), the optimization of function used for property prediction, and the effect of d[g(σ)]/d(P) vs P on the stability of steel properties based on the improved LMP method. All the data sets for the 9%—12%Cr steels are from NIMS database for the related calculations and analyses. The results show that the C value in LMP is not only different from steel to steel type but also varies with the multi–region stress levels, and the new approach to rupture life prediction proposes procedures for extrapolations of the short–term results, with rupture time measurements from tests lasting up to only 5×10³ h providing reasonable estimates of 10⁵ h rupture strengths, as well as the variation tendency of d[f(σ)]/d(P) vs P can reflect directly the long–term property stability of the steels investigated. Therefore, the concepts/methods proposed could improve effectively the accordance of predicted property values with observed ones and overcome obviously the overestimation tendency of 10⁵ h strengths, which are more suitable and easily realized to assess the long–term creep–rupture properties of the advanced high Cr ferritic steels.