← Back to HomeInternal

Data Dictionary

This page documents all data points and metrics collected by Gridiron Edge. Understanding the raw ingredients helps users interpret model outputs and understand the system's capabilities.

Play-by-Play Metrics (CFBD)

Efficiency Metrics

  • EPA (Expected Points Added): Offensive and defensive EPA per play. Measures the value created (or prevented) on each play relative to baseline expectations.
  • PPA (Points Per Attempt): Predicted points added per rushing or passing attempt. Measures efficiency by play type (rush PPA, pass PPA). Used in V2 unit grade calculations.
  • Success Rate: Percentage of plays with positive EPA. Tracks consistency and play-level effectiveness.
  • IsoPPP (Isolated Points Per Play): Average EPA on successful plays. Measures explosiveness and big-play ability.
  • Points Per Opportunity (PPO): Scoring efficiency when in scoring position. Tracks finishing ability.

Line & Havoc Metrics

  • Line Yards: Offensive line performance metric. Measures yards gained before contact.
  • Stuff Rate: Percentage of runs stopped at or behind the line of scrimmage.
  • Power Success: Conversion rate on short-yardage situations (3rd/4th & 2 or less).
  • Havoc Rate: Percentage of plays with TFL, INT, or PBU. Measures defensive disruption.
  • Havoc by Position: Front-7 havoc (linebackers/DL) and defensive back havoc tracked separately.

Play Type Splits

  • Run EPA / Pass EPA: Efficiency by play type (offensive and defensive).
  • Run Success Rate / Pass Success Rate: Consistency by play type.
  • Early Down EPA / Late Down EPA: Performance on 1st/2nd down vs. 3rd/4th down.

Field Position

  • Average Starting Field Position: Typical starting field position for drives. Impacts scoring opportunities.

V2 Unit Grades (Derived)

Unit-specific performance grades calculated from the metrics above, normalized to Z-scores:

  • Run Offense Grade: 50% Line Yards Z-score + 50% Rush PPA Z-score. Measures ground game effectiveness.
  • Run Defense Grade: 50% Stuff Rate Z-score + 50% (inverted) Rush PPA Allowed Z-score. Measures ability to stop the run.
  • Pass Offense Grade: 50% Pass PPA Z-score + 50% Pass Success Rate Z-score. Measures aerial attack effectiveness.
  • Pass Defense Grade: 50% (inverted) Pass PPA Allowed Z-score + 50% (inverted) Pass Success Rate Allowed Z-score. Measures pass defense strength.
  • Offensive Explosiveness: IsoPPP Z-score. Measures big-play ability.
  • Defensive Explosiveness: (Inverted) IsoPPP Allowed Z-score. Measures ability to prevent big plays.
  • Havoc Grade: Havoc Rate Z-score (season-level). Measures defensive disruption and playmaking ability.

Note: All grades are normalized to Z-scores (standard deviations from FBS mean), enabling direct comparison across different metric types. These grades power the "Unit Matchup" analysis on game detail pages.

Team Information

Talent & Recruiting

  • Talent Composite (247Sports): Overall team talent rating (0-100 scale). Aggregates recruiting rankings and roster quality.
  • Blue Chip Percentage: Percentage of roster with 4-star or 5-star recruits.
  • Star Counts: Number of 5-star, 4-star, and 3-star recruits on roster.
  • Recruiting Class Rankings: National and conference rankings for recent recruiting classes.

Team Classification

  • Conference: Conference affiliation (SEC, Big Ten, etc.).
  • Division: Division within conference (if applicable).
  • FBS Status: FBS (Football Bowl Subdivision) vs. FCS classification.

Returning Production

  • Offensive Returning Production: Percentage of offensive production (yards, TDs) returning from previous season.
  • Defensive Returning Production: Percentage of defensive production (tackles, INTs) returning from previous season.

Market Data (OddsAPI)

Spread Lines

  • Opening Line: Initial spread posted by sportsbooks.
  • Closing Line: Final spread at game time (or consensus closing line).
  • Line Movement: Tracked via timestamped line values to identify market shifts.
  • Multiple Books: Data from DraftKings, FanDuel, Caesars, BetMGM, and others.
  • Team Association: Each spread line is associated with a specific team (favorite or underdog).

Total Lines (Over/Under)

  • Opening Total: Initial over/under posted by sportsbooks.
  • Closing Total: Final total at game time.
  • Line Movement: Tracked to identify market consensus shifts.

Moneyline Odds

  • American Odds Format: Stored as positive (underdog) or negative (favorite) values (e.g., +150, -175).
  • Implied Probability: Calculated from odds using standard formulas:
    • Negative odds: |odds| / (|odds| + 100)
    • Positive odds: 100 / (odds + 100)
  • Favorite vs. Underdog: Each moneyline is associated with a specific team.

Market Metadata

  • Source Book: Sportsbook name (DraftKings, FanDuel, etc.).
  • Timestamp: When the line was collected (enables line movement tracking).
  • Consensus: Median/average across multiple books to identify market consensus.

Contextual Data

Weather Conditions

  • Temperature: Game-time temperature (affects passing efficiency and player performance).
  • Wind Speed: Wind conditions (impacts passing and kicking).
  • Precipitation Probability: Chance of rain/snow (affects ball handling and field conditions).
  • Condition Text: Human-readable weather description.
  • Source: Visual Crossing Weather API (historical and forecast data).

Injury Reports

  • Player Name: Injured player identifier.
  • Position: Player position (QB, RB, WR, OL, DL, LB, DB, etc.).
  • Severity: Injury status (Out, Questionable, Probable).
  • Team Association: Which team the player belongs to.
  • Note: Currently collected but not yet integrated into model adjustments.

Game Context

  • Venue: Stadium name and location.
  • Neutral Site: Boolean flag indicating if game is at a neutral location.
  • Home Field Advantage: Applied as +2.0 points for home teams (0.0 for neutral sites).
  • Kickoff Time: Game start time (used for timezone conversions and scheduling).

Derived Metrics

Power Ratings

Composite ratings calculated from the four pillars (Talent, Efficiency, Scoring, Results). See Methodology for details.

  • V1 Power Rating: Balanced composite (25% each pillar), normalized to Z-scores, scaled by 14.0.
  • Rating Format: "Points above average" (e.g., +15.0 means team is 15 points better than average).

Model Spreads & Totals

  • Model Spread: Calculated as (Home Rating - Away Rating) + HFA.
  • Model Total: Derived from offensive/defensive ratings, adjusted for pace and scoring efficiency.
  • Edge: Difference between model prediction and market consensus.

Win Probabilities

  • Model Win Probability: Derived from spread using sigmoid function: prob = 1 / (1 + 10^(spread / 14.5)).
  • Market Implied Probability: Calculated from moneyline odds using standard formulas.
  • Value: Difference between model probability and market implied probability (for moneyline bets).

Data Collection Status

Coverage Notes

  • Play-by-Play Metrics: 100% coverage for FBS teams (season-level) and FBS-vs-FBS games (game-level).
  • Market Data: Coverage varies by book and game importance. Major games have full coverage; smaller games may have limited book coverage.
  • Weather: Collected for all games, but forecast accuracy varies by proximity to game time.
  • Injuries: Collected but not yet integrated into model adjustments (future enhancement).
  • Historical Data: Limited by API tier access. Current season (2025) has full coverage; historical seasons may have gaps.

Data Sources

College Football Data (CFBD)

  • Play-by-play efficiency metrics (EPA, Success Rate, IsoPPP)
  • Team talent composite and recruiting data
  • Game schedules, scores, and team classifications
  • Returning production data

The Odds API

  • Spread, total, and moneyline odds from multiple sportsbooks
  • Opening and closing line tracking
  • Real-time odds updates (3x daily during active weeks)

Visual Crossing Weather

  • Game-time weather conditions (temperature, wind, precipitation)
  • Historical weather patterns