Current models for predicting the embrittlement of reactor pressure vessels are often developed on data derived from a single country in order to suit that country's requirements. Such models tend to predict data from a reactor fleet operating in another country more poorly. Previous investigations have found a possible correlation with initial yield stress. This investigation looks at the possibility of general improvements to models by the better use of data and incorporating yield stress in the predictive modeling. Previous models predict embrittlement from Charpy transition temperature shifts alone. This is an inefficient use of the underlying data and disproportionately propagates errors in the beginning-of-life transition temperature to all subsequent measurements. In this work we utilize all of the Charpy impact measurements. Three linear models have been developed to describe the upper shelf energy (USE), the temperature at the midpoint of the Charpy curve (tmid), and the inverse slope at tmid, S, in steels with less than 0.1% copper. These are then combined to predict full Charpy curves. The models produced good predictions of the USE and tmid but a poorer prediction of the S parameter; further optimization of this model is required in the future. When the three models were combined to predict full Charpy curves they gave reasonable predictions, indicating that this approach is viable. Charpy curves from different countries were predicted similarly well, suggesting that (i) this approach may better capture the influence of material and environmental variables on embrittlement and (ii) the yield stress is an important predictor variable.