An expanded methodology for testing how models behave under shifted or noisy data, missing values and different market regimes was adopted as standing policy for future research. It does not itself change any current result — it raises the bar that future work will be measured against.