Firstly, I want to congratulate you for the work you put in in the creation of ORE.
My question is concerned with the calculation of PFE. On the page 141 of the user guide you describe the method of calculation of PFE being the quantile of the NPV distribution. Am I being right in saying that this implies that your PFE is calculated not on the distribution of NPV in the real-world measure (i.e. in the traditionally denoted P-measure) but on the risk-neutral measure Q – which is not correct way to do it? If the answer is negative, can you point me to which method are you using to change from P-measure to Q-measure? In the book that you are referencing in the user guide for the cross-asset scenario generation (Modern Derivatives pricing and Credit Exposure Analysis) there are three methods which are described (traditional, risk-neutral adjusted and joint-measure model approach, Lichters, Stamm, Gallagher), do you follow one of these?
The ORE examples all preform a Risk-Neutral model calibration and thus the generated NPV distribution and PFE are indeed Risk-Neutral, this is the “traditional approach” from the above reference.
If you want to do a Real World calculation, and you are happy to use the RFE models in ORE, you can of course specify your own parameters directly (calibrated externally from historical data as you see fit, thus by-passing the Risk-Neutral calibration) and optionally disable mean reversion.