Football Data
Pipeline
The model only becomes as trustworthy as the data underneath it. PatternPicks separates schedule/results, team performance, player usage, market pricing and contextual signals before a recommendation can be published.
Games & Results
Season, week, kickoff, home/away, final score, game status and opponent. This becomes the canonical game key for every downstream record.
Team Performance
Scoring, efficiency, pace, turnovers, rushing/passing production, opponent-adjusted form and situational splits.
Player Usage
Attempts, carries, targets, receptions, yardage, touchdowns, red-zone work, goal-line work, snaps and routes where reliably sourced.
Market & Context
Sportsbook, exact line, odds, opening/current price, injuries, weather and other timestamped pregame context.
Frozen recommendation schema
| Group | Required fields |
|---|---|
| Identity | pick_id · game_id · market · player/team · model_version |
| Recommendation | side · exact_line · odds · sportsbook · maximum_playable_line |
| Model | projection · model_probability when calculated · market_implied_probability · estimated_edge |
| Evidence | supporting_signals · risk_signals · data_snapshot |
| Freeze | published_at · kickoff_at · immutable selection snapshot |
| Result | final_value · HIT/MISS/PUSH · profit/loss · closing_line_value where available |
Model families
Moneyline, spread, total and team-total projections.
Passing, rushing and receiving distributions with alternate-line evaluation.
Touchdown opportunity from usage, red-zone role, matchup and projected scoring environment.
Compares model output with price and rejects markets that move beyond the maximum playable threshold.
No pick without a complete snapshot
A recommendation cannot enter the verified ledger unless its game, market line, odds, source timestamp and model version are frozen before kickoff. Model probability and edge fields remain blank unless calculated by the active model.