Nang explains why traditional machine-learning pitfalls are amplified in finance: limited data, regime shifts, and the temptation to over-customise models for specific assets.
“Financial markets provide a small, noisy data set that makes overfitting especially dangerous”