How BOA earns trust
BOA separates data ingestion, model calculation, validation, publication, and postgame grading so a pipeline run alone never becomes a public claim.
Data provenance
Published football data is tied to known source records, stable IDs, season context, and update timestamps.
Certification before publication
Prediction and ranking updates must pass semantic and integrity checks before an admin can activate the next publication.
Forecast locking
Published forecast versions are treated as immutable historical artifacts. Corrections create a new version rather than silently rewriting the past.
Permanent model record
Completed forecasts remain visible with winner accuracy, margin error, Brier score, and projected-vs-actual results.
Protected admin writes
Public entities are read-only to visitors. Publishing, syncing, and model mutations remain restricted to administrative paths.
Corrections policy
Data errors should be corrected at the source or mapping layer, documented, revalidated, and republished through a new traceable artifact when necessary.
What BOA measures—and what it does not
- • Team ratings and national ranking movement
- • Pregame win probabilities, projected scores, margins, and BOA spreads
- • Data-driven player grades and verified statistics
- • Live CFBD-backed game-state intelligence and preserved snapshots
- • Recruiting class and position-group data inputs
- • Film-charting equivalence to professional scouting grades
- • Injury-adjusted probabilities without a validated injury model
- • Starter or replacement-risk probabilities without a validated roster model
- • A BOA-native play-by-play win-probability model before it is tested
- • Hypothetical simulator outputs as actual rerun model forecasts