Cohort Differentiation (Backward Analysis): Practical Implementation
How to perform backward (diagnostic) analysis in conditional trading studies — comparing feature distributions across success and failure cohorts, quantifying their separation, and turning those insights into context-aware refinements.