Chapter 12: Key Takeaways

Quick Reference Guide for Defensive Metrics and Analysis

Core Defensive Statistics

Metric Definitions

Interpretation Guidelines

High tackles can indicate: - Aggressive defensive style (positive) - Compensation for poor positioning (negative) - High involvement due to team style (contextual)

Low tackles can indicate: - Excellent positioning deterring challenges (positive) - Playing for dominant possession team (contextual) - Avoidance of engagement (negative)

Key Formulas

Possession-Adjusted Defensive Actions (PADA)

$$\text{PADA} = \frac{\text{Defensive Actions per 90}}{1 - \text{Team Possession \%}}$$

Example: - 2.5 tackles per 90, team possession 65% - PADA = 2.5 / 0.35 = 9.14 tackles per opponent possession

Passes Per Defensive Action (PPDA)

$$\text{PPDA} = \frac{\text{Opponent Passes (Their Def Third)}}{\text{Defensive Actions (Their Def Third)}}$$

Benchmark Values: | Style | PPDA | |-------|------| | High Press | < 8 | | Medium Press | 8-12 | | Medium Block | 12-15 | | Low Block | > 15 |

Tackle Success Rate

$$\text{Success Rate} = \frac{\text{Tackles Won}}{\text{Total Tackles}} \times 100$$

Aerial Dominance Index

$$\text{ADI} = (\text{Aerial Win Rate} - 0.5) \times \text{Total Aerial Duels}$$

Pressing Metrics Summary

Individual Pressing

Team Pressing

Counter-Pressing

Defensive Value Calculations

xG Prevented Components

1. Shot Blocks: Direct xG of blocked shots
2. Interceptions: xT at intended destination
3. Tackles Won: Positional value × 0.02-0.05
4. Clearances: Danger zone value × 0.01-0.04
5. Pressure Regains: xT at location × 0.5

Defensive xT Prevented Formula

$$\text{xT Prevented} = \sum_{i} \text{xT}(\text{intercept\_location}_i)$$

Center-Back Archetypes

Contextual Adjustment Checklist

Always Adjust For:

1. Team Possession % - Use PADA formula
2. Opposition Strength - Weight by opponent xG created
3. Game State - Segment by leading/level/trailing
4. Position - Compare within position groups

When Interpreting:

• Consider team tactical instructions
• Account for partner profiles
• Note sample size limitations
• Examine underlying context

Visualization Best Practices

Defensive Action Maps

• Use pitch diagram with scatter plots
• Color-code by action type
• Size by success/value

Radar Charts

• Include 7-8 key metrics
• Normalize to 0-100 scale
• Add comparison overlays

Defensive Shape

• Use convex hull for team shape
• Show average positions
• Include positional spread metrics

Practical Application Guide

Defender Evaluation Checklist

1. Core Stats: Tackles, interceptions, clearances, blocks
2. Aerial Ability: Win rate and involvement
3. Pressing: Pressure frequency and success
4. Ball-Playing: Pass completion and progression
5. Context: Team, opposition, game state adjustments
6. Value: xG prevented, defensive xT

Recruitment Criteria by Role

Ball-Playing CB Requirements: - Pass completion > 88% - Progressive passes > 3.5 per 90 - Comfortable under pressure

Aerial Stopper Requirements: - Aerial win rate > 70% - Clearances > 7.0 per 90 - Heading accuracy in both boxes

Pressing CB Requirements: - Pressures > 15 per 90 - Tackles won > 2.0 per 90 - Recovery pace for high line

Common Mistakes to Avoid

1. Comparing raw stats without possession adjustment
2. Ignoring tactical context in interpretation
3. Overvaluing volume over efficiency
4. Single-metric evaluation instead of profiles
5. Ignoring deterrence effect (what doesn't happen)

Quick Reference: Metric Benchmarks

Elite Level (Top 5%)

Good Level (Top 25%)

Key Takeaways Summary

1. Defensive analysis requires context - Raw stats are misleading without adjustment

2. No single metric captures defensive quality - Use multi-dimensional profiles

3. The counterfactual problem - Great defenders prevent actions that never occur

4. Pressing is measurable - PPDA, high turnovers, and counter-pressing quantify pressing effectiveness

5. Archetypes matter - Different defenders suit different systems

6. Value can be estimated - xG prevented and defensive xT provide value frameworks

7. Team defense emerges from individuals - Analyze both levels

Code Snippets

Basic Defensive Stats

def get_defensive_stats(events_df, player_name, minutes):
    player = events_df[events_df['player'] == player_name]
    p90 = 90 / minutes

    return {
        'tackles_p90': len(player[player['type'] == 'Tackle']) * p90,
        'interceptions_p90': len(player[player['type'] == 'Interception']) * p90,
        'clearances_p90': len(player[player['type'] == 'Clearance']) * p90
    }

PPDA Calculation

def calculate_ppda(events_df, pressing_team):
    opponent = [t for t in events_df['team'].unique() if t != pressing_team][0]

    opp_passes = events_df[
        (events_df['team'] == opponent) &
        (events_df['type'] == 'Pass') &
        (events_df['location'].apply(lambda x: x[0] < 40 if isinstance(x, list) else False))
    ]

    def_actions = events_df[
        (events_df['team'] == pressing_team) &
        (events_df['type'].isin(['Pressure', 'Tackle', 'Interception'])) &
        (events_df['location'].apply(lambda x: x[0] > 80 if isinstance(x, list) else False))
    ]

    return len(opp_passes) / len(def_actions) if len(def_actions) > 0 else float('inf')

Possession Adjustment

def possession_adjust(actions_p90, team_possession):
    return actions_p90 / (1 - team_possession)