Current hypothesis
Roster grade, rank tier, units, staff and continuity contain correlated judgments.
Evidence and counterevidence
Tech received +3.5 roster, +1 skill, +1.5 trenches, +1 staff and +1.5 integration points before quarterback and venue.
Counterevidence
Notre Dame’s heavily stacked advantages understated its winning margin; removing points is not automatically better.
Unknown
The independent predictive contribution of each component is not established.
Next diagnostic action
Build an evidence-to-component map and a simpler baseline; compare on future held-out games before adoption.
What would support it
The hypothesis strengthens when multiple adjustments depend on the same fact and add no independent predictive benefit in later tests.
What would weaken or invalidate it
It weakens when components have distinct evidence and improve held-out forecasts beyond a simpler baseline.