Key Takeaways: Special Teams Analytics

One-page reference for Chapter 11 concepts

Special Teams Impact

Kicker Evaluation

# Expected FG% by distance
def expected_fg(distance):
    if distance <= 20: return 0.99
    elif distance <= 30: return 0.95
    elif distance <= 40: return 0.88
    elif distance <= 50: return 0.75
    elif distance <= 55: return 0.60
    else: return 0.45

# FG Over Expected
fg_over_expected = actual_makes - sum(expected_fg(d) for d in distances)

FG% by Distance Benchmarks

Punter Evaluation

# Net average
net_avg = gross_yards - return_yards

# Inside 20 rate
inside_20 = (ending_position <= 20).mean()

# Better: EPA per punt
punt_epa = punts['epa'].mean()

Punting Benchmarks

Return Evaluation

# Kick return value
kr_avg = kickoff_returns['return_yards'].mean()
kr_value = kr_avg * 0.04  # EP per yard

# Punt return value
pr_avg = punt_returns['return_yards'].mean()

# Yards over expected
yoe = actual_yards - (attempts * league_avg)

Return Benchmarks

Coverage Quality

# Coverage score (higher = better)
coverage_score = (
    fair_catch_rate * 100 -
    long_return_rate * 100 -
    avg_return_allowed * 2
)

Field Position Value

Every 10 yards ≈ 0.4 EP

Decision Framework

def should_go_for_it(yardline, ytg):
    """Simplified 4th down decision."""
    conv_prob = 0.75 - ytg * 0.05
    go_ev = conv_prob * 2.5 + (1-conv_prob) * (-field_position_value(yardline))
    punt_ev = -field_position_value(yardline - 40)
    return go_ev > punt_ev

Sample Size Warning

Small samples = wide confidence intervals

Year-to-Year Stability

Lower correlation = higher variance/luck

Team ST Evaluation Framework

1. Kicking
   - FG over expected
   - XP%
   - Kickoff touchback rate

2. Punting
   - Net average
   - Inside 20 rate
   - Hangtime proxy (fair catch rate)

3. Returns
   - KR average
   - PR average
   - Return TDs

4. Coverage
   - KR allowed
   - PR allowed
   - Return TDs allowed

Common Pitfalls

Data Sources

Key Limitations

• Sample sizes too small for confidence
• Environment effects hard to model
• Unit coordination not captured
• Year-to-year stability moderate
• High-leverage context difficult

Preview: Part 3

Next: Team Analytics - how individual performances combine into team success and what drives winning.