Concepts
The vocabulary of cohort intelligence.
Four canonical pages defining the methodology terms Chart Library is built on. Read the pillar (cohort intelligence) first if you’re new; the rest go deeper on calibration and eval rigor.
Cohort intelligence →
The pillar concept. The alternative to point-prediction stock forecasting — return the cohort of historical analogs, the full distribution of what they did, and the features that separated winners from losers.
Calibrated stock forecasting →
What calibration means and why distributions beat point estimates. How we validate that 80% bands actually contain 80% of held-out outcomes.
Symbol-disjoint evaluation →
The eval discipline that prevents leakage in financial ML. Why date-based train/test splits silently inflate results 5-30 percentage points and how to do it honestly.
Conformal prediction in finance →
The math behind the calibrated probability bands. Method-agnostic, distribution-free correction layer that turns any regression model's bands into properly calibrated bands.