Chapter 60

Capstone - Complete Analytics System

Intermediate 30 min read 5 sections 10 code examples
0 of 60 chapters completed (0%)

15.1 Understanding Team Performance

Team-level analytics aggregate individual performances and tactical patterns into holistic assessments of team quality. This chapter explores how to measure team performance, identify playing styles, and predict future results using statistical models.

Learning Objectives
  • Understand team-level xG and its applications
  • Analyze team playing styles using aggregate metrics
  • Build expected points models
  • Evaluate defensive and attacking team quality
  • Create team comparison dashboards

The Three Pillars of Team Analysis

Chance Quality

How good are the chances created and conceded? (xG for/against)

Chance Volume

How many chances created and conceded? (shots for/against)

Conversion

How efficiently are chances converted? (Goals vs xG)

15.2 Team xG Analysis

Team-level expected goals (xG) provides a more stable measure of performance than actual goals, which are subject to variance. Key team xG metrics:

Metric Definition What It Indicates
xG For (xGF) Total xG created by the team Attacking quality
xG Against (xGA) Total xG conceded by the team Defensive quality
xG Difference (xGD) xGF - xGA Overall team quality
xGD per 90 xGD normalized per match Sustainable performance level
Goals - xG Actual goals minus expected Finishing luck/skill

xG Trend Analysis

Tracking xG over time reveals performance trajectories and potential regression:

15.3 Expected Points (xPts)

Expected points models predict how many points a team should earn based on their xG performance. This helps identify over/underperforming teams.

Expected Points Calculation

Using Monte Carlo simulation or Poisson distribution, we can estimate the probability of each match outcome (win/draw/loss) based on xG, then convert to expected points.

League Table Based on xPts

15.4 Team Playing Style Analysis

Aggregate metrics can reveal distinct team playing styles. Key dimensions include:

Possession Style
  • Possession %
  • Passes per 90
  • Pass completion %
  • PPDA (pressing intensity)
Directness
  • Long ball %
  • Progressive passes
  • Counter-attacks
  • Avg. sequence length
Defensive Approach
  • High press %
  • PPDA
  • Defensive line height
  • Tackles in def/mid/att 3rd

15.5 Attack vs Defense Quality

Plotting attacking quality against defensive quality reveals team strengths and weaknesses:

15.6 Team Radar Charts

Radar charts effectively compare team profiles across multiple dimensions:

15.7 Predictive Modeling

Team-level metrics can be used to build predictive models for future performance:

Regression to the Mean

Teams that significantly over/underperform their xG tend to regress. xG-based predictions are often more accurate than using actual goals.

15.8 Practice Exercises

Exercise 15.1: Expected Points Table Analysis

Task: Create a full league table based on expected points. Compare to actual standings and identify the biggest over/underperformers. Create a dumbbell plot showing actual vs expected position changes.

Exercise 15.2: Playing Style Clustering

Task: Using possession, PPDA, long ball %, and crosses per 90, cluster teams into playing style groups. Use K-means to identify 4 distinct styles and create a PCA visualization with style labels.

Exercise 15.3: Season Points Prediction Model

Task: Build a model to predict end-of-season points using early-season xG difference data. Test how predictive power (R-squared) improves as more games are included (5, 10, 15, 20 games). Visualize the increasing predictive power.

15.9 Chapter Summary

Key Takeaways
  • Team xG provides more stable performance measurement than actual goals
  • Expected points reveal true team quality beyond points actually earned
  • Playing style analysis uses aggregate metrics to identify tactical patterns
  • Attack/Defense quadrant plots quickly identify team strengths and weaknesses
  • Regression to mean is a key concept - teams tend to move toward their xG-based performance
  • Predictive models using xG often outperform those using actual goals
Congratulations!

You've completed the foundational chapters of Soccer Analytics! You now have the tools to analyze players and teams using modern statistical methods. Continue exploring advanced topics like tracking data, player valuations, and machine learning applications in football analytics.