Accuracy trust
Hit-rate methodology
Homefield publishes rates only on *your* Pro ledger after enough scored decisions — never a sitewide “most accurate” claim.
What counts
| Metric | Denominator (n) | Definition |
|---|---|---|
| Agreement | Locked decisions with a user pick | Share of locks where your pick IDs matched Homefield’s recommendation. |
| HF hit rate | Scored disagreements only | On scored *disagreements* only: share where Homefield’s locked side outscored yours (actuals). |
| Followed hit | Scored agreements | On scored *agreements*: share where the shared pick won (or tied) on actuals. |
Publish gate: n ≥ 30 for that metric. Until then the UI shows n=X (need 30) — never a percent.
Rules we publish against
- Publish gate: n ≥ 30 for that metric’s own denominator (locks, scored disagreements, or scored follows).
- Sample = locked decisions with a user pick. HF hit rate uses scored disagreements only — agreeing with Homefield does not inflate that rate.
- Baseline: coin-flip 50% on disagreements. Beating chalk/experts is not claimed.
- Privacy: per-account Rivalry desk only. No public leaderboard of hit rates without an explicit opt-in later.
- Anti-claim: never “most accurate projections,” “#1,” or “beat FantasyPros.”
FAQ
When does Homefield publish a hit-rate percent?
Only after that metric’s own denominator reaches n ≥ 30. Until then the Rivalry desk shows n=X (need 30) — never a padded percent.
What is Homefield’s accuracy baseline?
Coin-flip 50% on scored disagreements. Homefield does not claim to beat chalk ADP, FantasyPros ECR, or experts.
Is there a public Homefield hit-rate leaderboard?
No. Rates live on your Pro Rivalry ledger only. A sitewide aggregate would need an explicit privacy opt-in later — never marketed as “most accurate.”
Analysis software — not betting advice. Homefield advises; your host places. Canonical helpers live in src/lib/accuracy/publish.ts · desk: /rivalry.