Confidence reflects signal coverage, ML depth, conflict penalties, and historical cohort precision when enough labels exist.
Confidence is about coverage and historical trust, not bullishness. Thin or conflicted inputs lower it.
A high Breakout Score™ with low confidence deserves more skepticism than a slightly lower score with deep coverage.