Key Takeaways: Home Field Advantage

One-page reference for Chapter 15 concepts

HFA Measurement Methods

# Win percentage method
home_win_pct = home_wins / total_games

# Point margin method
hfa_points = avg_home_score - avg_away_score

# Spread method (market-implied)
implied_hfa = -avg_spread_line  # ~2.5-3.0 points historically

Historical HFA Values

Era Home Win % Implied Points
2000-2010 57-58% ~3.0
2011-2019 54-57% ~2.5-3.0
2020 (COVID) 50-51% ~0.5-1.0
2021-2023 52-54% ~2.0-2.5

HFA Trend

# Declining over time
# Historical: ~3.0 points
# Current: ~2.0-2.5 points

# COVID impact showed crowd = major factor
# 2020 home win % ≈ 50% (no crowds)

Causal Factors

Factor Impact Evidence
Crowd Noise High False start differential
Travel Medium West coast teams worse traveling east
Time Zones Low-Medium Early games after timezone change
Referee Bias Low Slightly more away penalties
Familiarity Low Knowing venue, climate

Team-Specific HFA

Team Factors HFA
SEA Noise, weather High (~3.5 pts)
KC Arrowhead noise High (~3.5 pts)
DEN Altitude (5,280 ft) High (~3.0 pts)
GB Cold, Lambeau Above Avg (~3.0 pts)
NO Dome noise Above Avg (~3.0 pts)

Penalty Differentials

# False starts (crowd noise effect)
away_false_starts > home_false_starts  # ~15-20% more

# Delay of game (communication issues)
away_delays > home_delays  # Similar pattern

Travel Effects

Distance Away Win % Impact
< 500 mi Minimal
500-1000 mi Slight decrease
1000-2000 mi Small decrease
> 2000 mi Noticeable (~2-3%)

Timezone Effects

# West to East travel (worst)
away_win_pct ≈ 43-45%

# East to West travel (better)
away_win_pct ≈ 46-48%

# Same timezone
away_win_pct ≈ 47-48%

Using HFA in Predictions

# Basic prediction
predicted_spread = (home_rating - away_rating) + hfa

# Team-specific
predicted_spread = neutral_spread + get_team_hfa(home_team)

# Default HFA values
standard_hfa = 2.5  # Modern NFL
team_specific_range = 1.5 to 3.5  # Varies by venue

Model Integration

# Power rating approach
expected_margin = home_power - away_power + hfa

# Win probability
home_win_prob = sigmoid(expected_margin / 13.5)

# Example
# KC (rating +4) hosts LV (rating -2)
# HFA = 3.0 (Arrowhead)
# Predicted: KC by 9.0 points

Key Correlations

Relationship Finding
HFA vs Wins r ≈ 0.15-0.20 (small)
Crowd Size vs HFA Positive (COVID evidence)
Altitude vs HFA Positive (Denver effect)
Travel Distance vs Away Performance Negative (small)

Practical Applications

Use Case HFA Approach
Spread prediction Add 2.0-2.5 points
Power rankings Neutral-site adjustments
Playoff seeding value Home = 2-3 point swing
Model building Include as fixed parameter

Key Insights

  1. HFA has declined from ~3 to ~2-2.5 points
  2. Crowd support is primary driver (COVID evidence)
  3. Team-specific values improve predictions
  4. Travel/timezone effects are small but real
  5. Playoff home field still valuable (~2-3 pts per round)

Preview: Chapter 16

Next: Strength of Schedule - measuring and adjusting for opponent quality.