nfl-prediction-data

Name: fbdata.xyz
Creator: fbdata
License: https://creativecommons.org/licenses/by/4.0/

Workflow map and transparent baseline for agents building NFL forecasts.

[CONTRACT]

product: point-in-time NFL data plus a reproducible baseline model
primary risk: lookahead bias from upstream refits; mitigate with as_of
latest: allowed for current scouting, unsafe for historical backtests
snapshot_window: 2026-08-03T20:45:37.669938Z to 2026-08-10T15:43:58.470767Z; earlier as_of returns empty by design
pricing: returned as x-fbdata-price-usd; settlement via x402 gateway

[WORKFLOWS]

spread_model: teams + adjusted + games
total_model: teams + tendencies + games
fantasy_projection: passing + rushing + receiving
player_props: receiving + rushing + pressure + coverage
single_game_context: /v1/agent/context?game_id=...
baseline_predictions: /v1/predictions/games?season=...&week=...
raw_features: /v1/schema/pbp then /v1/query with explicit columns

[RECOMMENDED_CALLS]

GET /v1/leaders/adjusted?season=2025&format=json team power features: adjusted offense, defense, net and schedule swing

GET /v1/leaders/games?season=2025&format=csv closing spread/total, final result, ATS and over/under resolution

GET /v1/leaders/receiving?season=2025&week_min=1&week_max=9&min=40&as_of=2025-11-03T00:00:00Z example of a week-10 backtest-safe receiving feature slice

GET /v1/schema/pbp choose observed vs derived features before training

GET /v1/predictions/games?season=2025&week=1 versioned EPA/scoring baseline with fair spread, total, probabilities and market edges

GET /v1/agent/context?game_id=2025_01_DAL_PHI one decision packet: game, model, market, props, availability, weather and missing feeds

[GUARDRAILS]

The built-in model is an auditable baseline, not a guarantee or wagering instruction. For any historical experiment, require x-fbdata-point-in-time: true and confirm the requested date is inside /v1/catalog snapshot_window, or fail closed.