Regression Analysis

Identifying Unsustainable Performance in NFL Data

What is Regression to the Mean?

Regression to the mean is the statistical phenomenon where extreme performances tend to move back toward average over time. In NFL betting, identifying regression candidates creates significant edges.

The Core Concept

Some statistics are highly variable (high luck component), while others are stable (skill-based). Luck-driven outliers will regress; skill-driven outliers persist.

Why Regression Matters for Betting
  • Markets overweight recent performance - Public bettors chase hot teams
  • Unsustainable stats inflate/deflate lines - Creates value opportunities
  • Sharp bettors exploit regression - Understanding it is essential

High Regression Statistics (Luck-Driven)

These stats have low year-over-year correlation and high variance. Extreme values will likely regress:

Why It Regresses
  • Fumble recovery is ~50/50 (pure luck)
  • INT rates fluctuate based on randomness
  • Year-over-year correlation: ~0.20
Betting Application
  • Fade teams with extreme positive differential
  • Buy teams with extreme negative differential
  • Don't trust early-season TO leaders
Example: A team +12 in turnover differential through 8 games is not "great at creating turnovers" - they're running hot and will regress.

Why It Regresses
  • NFL average: ~55-58% TD rate
  • Year-over-year correlation: ~0.25
  • Small sample size amplifies variance
Betting Application
  • Team at 75%+ RZ TD rate will regress down
  • Team at 40% or below will regress up
  • Affects point differential interpretation

Why It Regresses
  • NFL average: ~38-42%
  • Year-over-year correlation: ~0.30
  • Driven by opponent quality, randomness
Betting Application
  • Defense allowing 28% will regress up
  • Defense allowing 50%+ will regress down
  • Schedule strength impacts early numbers

Why It Regresses
  • Defenses have minimal impact on FG%
  • Kicker skill >> defensive pressure
  • Almost pure randomness
Betting Application
  • Teams "lucky" vs kickers will regress
  • Don't credit defense for missed FGs
  • Opponent FG% is noise

Stable Statistics (Skill-Driven)

These stats have high year-over-year correlation. Extreme values are more likely to persist:

Statistic YoY Correlation Why It's Stable
EPA per Play ~0.55-0.65 Reflects true offensive/defensive quality
Success Rate ~0.50-0.60 Measures consistent execution
Yards per Play ~0.45-0.55 Fundamental efficiency metric
Pass Block Win Rate ~0.50+ OL skill is consistent
Completion % Over Expected ~0.45-0.55 QB accuracy is a stable skill
Trust These Metrics

When a team has strong EPA, success rate, or CPOE, these indicate genuine quality. Unlike turnover differential, these metrics predict future performance.

Regression Estimator

Estimate where a stat will regress based on sample size and league average:

Common Examples:

Practical Betting Applications

1. Early Season Regression Plays

Scenario: Team starts 5-1, +10 turnover differential, 70% RZ TD rate

Market Perception: "This team is elite, one of the best in the league"

Reality: Running extremely hot on high-variance stats

Action: Fade this team against the spread, especially vs quality opponents. Their record is inflated by luck.

2. Unlucky Team Buys

Scenario: Team is 3-5, -8 turnover differential, strong EPA numbers

Market Perception: "Team can't close games, something wrong"

Reality: Fundamentally solid, experiencing bad luck

Action: Buy this team against the spread. Record will improve as turnovers regress.

3. Point Differential Adjustment

Key Adjustments:

  • Add/subtract ~2.5 points per turnover differential from expected (vs actual)
  • Adjust for red zone TD rate vs league average
  • Expected wins often differ significantly from actual wins early season

Regression Red Flags

Watch for these signs that a team is due for regression:

Regression Down (Overperforming)
  • Turnover differential > +8
  • Red zone TD rate > 70%
  • Opponent 3rd down rate < 32%
  • Record significantly > Pythagorean expected
  • Close game record > 70%
Regression Up (Underperforming)
  • Turnover differential < -8
  • Red zone TD rate < 45%
  • Opponent 3rd down rate > 48%
  • Record significantly < Pythagorean expected
  • Close game record < 30%
Key Takeaways
  • Turnovers are mostly luck - regression is certain
  • EPA and success rate are stable/predictive
  • Early-season records are often noise
  • Fade teams running hot on luck stats
  • Buy unlucky teams with good fundamentals
Stat Stability

High Regression (Luck):

  • Turnover differential
  • Red zone TD rate
  • Opponent FG%
  • Close game record

Low Regression (Skill):

  • EPA per play
  • Success rate
  • CPOE
  • Pass block win rate
Learn EPA Analysis

EPA is a stable, predictive metric - learn how to use it

EPA Tutorials