1 of 37

Football Data Analytics Portfolio

February 2024

EDD WEBSTER

2 of 37

PORTFOLIO OVERVIEW

This portfolio aims to demonstrate the professional data science experience that I have gained over the last five years, developed from working in an elite, first team football environment, as well as several best-in-class organisations.

A strategic approach to dissecting and addressing examples of difficult industry issues are discussed in detail. For each challenge, an approach and a summarised solution is provided with some real-life examples.

Finally, this portfolio aims to showcase the measurable impacts of the challenges tackled, as well as provide opportunity to demonstrate further work and skills.

The decks demonstrates a broad range of data insights skills, ranging from data processing, modelling, visualisation and results communication.

3 of 37

  • Introduction (slides 4-8)

  • Areas of Impact (slides 9-10)

    • Football Intelligence Reporting (slides 11-12)

    • Insight Delivery to Football Stakeholders (slides 13-16)

    • Internal Scouting and Recruitment Player Rating System (slides 17-21)

    • Data and Infrastructure Engineering (slides 22-32)

    • Open Source Contributions (slides 33-34)

  • Summary (slides 35-37)

CONTENTS

Contents

3

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

All publicly available code: github.com/eddwebster/football_analytics and Tableau visualisations: public.tableau.com/profile/edd.webster.

Summary

4 of 37

4

INTRODUCTION

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

5 of 37

The reason for producing this

football data analytics portfolio?

To demonstrate my broad range of data science and data engineering skills, as well as my experience and knowledge of football analytics concepts and analytical techniques.

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

5

Summary

6 of 37

INTRODUCTION

Timeline of My Background and Education Over the Last Ten Years

6

Data

Scientist

Apr 2022 – Apr 2023

London, UK

Business

Analyst

Aug 2018 – May 2018

London, UK

Lead

Data Analyst

Jul 2018 - Apr 2020

Lima, Peru

MSci Degree

Chemistry

2012 – 2016

London, UK

Team

Leader

Jan 2017 – Jul 2017

Tamale, Ghana

2016

2017

2018

2019

2020

2021

2022

2024

Senior Data & CRM

Manager

Oct 2021 – Apr 2022

London, UK

Data Science

Consultant

May 2022 – Apr 2023

Remote

Lead

Data Scientist

Apr 2023 – June 2024

Leicester, UK

Data & Insights

Analyst

Apr 2020 – Oct 2021

Manchester, UK

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

2023

Senior Data

Scientist

Jun 2024 – Present

Remote, UK

7 of 37

Objectives of the Data Insights Department of a Football Club

INTRODUCTION

7

To provide access to data for the club, to enable every decision made by for football operations to be objective and informed by data.

Educate decision makers and empower them to want to use data.

To maximise performance and subsequently, win more games.

1

2

3

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

8 of 37

Areas of a Football Club that Data Insights Can Have an Impact

INTRODUCTION

8

Recruitment and scouting – using models to bring a positive impact on player trading outcomes in the transfer market;

Performance evaluation and player development – measure the performance of players through models;

Sport science and athlete monitoring – provide reviews, audits, and recommendations for interventions; and

Club strategy – discovery of new insights through physics-based modelling that can be implemented by coaches in team strategy and affect performances on the pitch.

1

2

3

4

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

9 of 37

Data-Driven Decision Taking Place Across a Football Club Club

INTRODUCTION

9

Manager

Director of Football

Academy

Agents

Performance

Analysts

Data Scientists

Players

Data

Recruitment

and Scouts

10 of 37

Areas of Expertise Required in a Functioning Data Analytics Team

INTRODUCTION

10

Data

Engineer

Data

Scientist

Data Insights Analyst

Performance Analyst

Technical

Expert

Infrastructure

Organise

Support

Connect

Enrich

Explore

Discover

Predict

Investigate

Infom

Visualise

Advise

Bridge

Connect

Align

Insights

Observe

Interpret

Context

Implement

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

Slide recreated from a slide by Tom Garner in his presentation: “Delivering Football Performance Insights” at the StatsBomb Conference 2023: https://www.youtube.com/watch?v=ApWqVlz-ztQ&ab_channel=StatsBomb. Slide can be found at around 4m30s

11 of 37

11

AREAS OF IMPACT

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

12 of 37

Application of Data Across Different Workstreams

AREAS OF IMPACT

Football Intelligence Reporting

Data and Infrastructure Engineering

Insight Delivery to Football Stakeholders

Internal Scouting Player Rating System

Open Source Contributions

The application of data to make evidence-based decisions, that are delivered efficiently and effectively to seniors stakeholders and football practitioners.

Adding context to the scouting and recruitment process with data via a bespoke, player rating system, based on a system of defined and weighted metrics, custom-made to the needs of the First Team and the game model.

Creation of a back-end infrastructure, to enable automated reporting, player mapping, and predictive modelling, including the creation of the ‘Player 360’ reference tables.

Open source projects using publicly available event and tracking data. Projects include working with tracking data including Pitch Control modelling, Chance Quality modelling of shots from Event data, and determining ball-playing centre backs through clustering, to determine the ”the next Gerard Piqué

Automated reporting for opposition and post-match analysis, implementing bespoke KPIs and modelled metrics, curated to the football club’s game model. Custom reports for the purposes of recruitment, performance analysis and sports science.

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

13 of 37

13

FOOTBALL INTELLIGENCE REPORTING

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

14 of 37

Turning Big Data into Meaningful Insights Using Tableau

FOOTBALL INTELLIGENCE REPORTING

14

Simplicity - deliver clear and concise messages that can be comprehended quickly and concisely.

Enable Control - dashboard and visualisations need to include filters, tool tips, and actions, that can allow the user (often coaches) to find the insights themselves without the need for programming or dashboard building skills.

Flexibility - filters, tooltips and actions can allow the user to find the insights themselves.

1

2

3

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

15 of 37

A Selection of Football Insight Analysis Tools 1/2

FOOTBALL INTELLIGENCE REPORTING

15

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Tableau visualisations created with publicly available data can be found at my Tableau Public profile: public.tableau.com/profile/edd.webster.

Summary

16 of 37

A Selection of Football Insight Analysis Tools 2/2

FOOTBALL INTELLIGENCE REPORTING

16

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Tableau visualisations created with publicly available data can be found at my Tableau Public profile: public.tableau.com/profile/edd.webster.

Summary

17 of 37

17

INSIGHT DELIVERY TO FOOTBALL STAKEHOLDERS

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

18 of 37

Providing Post-Match and Opposition Analysis to Key Stakeholders

INSIGHT DELIVERY TO FOOTBALL STAKEHOLDERS

18

19 of 37

Delivering Insights Directly to the Club’s Senior Stakeholders

INSIGHT DELIVERY TO FOOTBALL STAKEHOLDERS

"Our Toughest Match" 🎙️ | Enzo Maresca Previews Rotherham United. Stats quoted 3 minutes and 58 seconds into the video: youtu.be/kWlyTNaqXCA?si=1HotJWAVe9TkehW1&t=238

"A Complicated Team" | Enzo Maresca Assesses Plymouth. Stats quoted 4 minutes and 6 seconds into the video: youtu.be/eRyX4LNp6LE?si=ZyGG9IWGM8suTugK&t=246

20 of 37

Delivering Insights as Part of Analytics FC Coach ID Service

INSIGHT DELIVERY TO FOOTBALL STAKEHOLDERS

20

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

21 of 37

Tailor-made Analysis for Online Publication

INSIGHT DELIVERY TO FOOTBALL STAKEHOLDERS

21

22 of 37

22

PLAYER RATINGS FRAMEWORK

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

23 of 37

The Purpose of Data Analytics for Recruitment and Scouting

PLAYER RATINGS FRAMEWORK

23

To narrow a whole cohort of players to a manageable number of targets for further investigation by domain experts, to optimise the resources available to the department.

Provide an objective assessment of player’s in-game performance for both their output and playing style.

Empower decision makers that want to use the data to further inform their decision making and increase the likelihood of success for the purposes of player recruitment.

1

2

3

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

24 of 37

24

How do you measure…

  • Goals
  • Shots
  • Shots on Target
  • Shot Conversion
  • Shot on Target Conversion
  • Pressured Shot
  • Range
  • xG

Shooting?

Scoring?

End Product?

Finishing?

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

25 of 37

  • PROSPECTS ratings: Performance Ratings for Objective Scouting of Players to Evaluate Contribution and Technical Skill.
  • A rating system designed with Leicester City’s game model in mind, bespoke to the qualities of players that football practitioners at the clubs such as the manager, coaching staff, and scouting and recruitment are looking for.
  • The PROSPECTS Data-Driven Player Rating System aims to provide accurate measurements of performance for a player in a match based on their in-game event data (passing, shooting, duels, pressing, etc.). These ratings can then be aggregated and determined for a specific period or whole season across a competition.
  • The ratings framework takes into account not just player output such as Value Framework outputs (xT, OBV, etc.) , xG, or success percentages (goal conversion, pass completion), it is also used to evaluated playing style.
  • The PROSPECTS Player Ratings takes into account over 200 different metrics derived from event data, with the final score being devised from six Categories, corresponding to different Technical Qualities for a player’s performance specific to the Leicester City game model. The six sub-scores are:
    • Scoring
    • Passing
    • Possession
    • Defending
    • Creating / Attacking
    • Goalkeeping
  • These six categories are then broken down further into Technical Qualities that are discussed further on the following slides.
  • Using each of the six categories, the Overall Ratings is calculated according to the player’s position in the match. This ensures that ratings are distributed fairly across each of the categories regardless of position. For example, a defender would not be expected to perform as well in the attacking category as a centre-forward, and therefore the weighting applied to defenders and centre-forwards for those actions are different.
  • These raw scores are then smoothed out to measure how close they are to the average (Z-scores) and then converted into an intuitive rating of 1-100 to provide a final rating.

PROSPECTS Rating System

PLAYER RATINGS FRAMEWORK

25

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

26 of 37

PROSPECTS Rating Proposed Categories and Respective Qualities

PLAYER RATINGS FRAMEWORK

26

Scoring

Creating

Passing

Goals

Shots

Conversion

Position

Crossing

Assists

Dribbles

Carries

Involvement

Accuracy

Direction

Intent

Composure

On-Ball Decisions

Progressive Passing

Interceptions

Possession

Defending

Goalkeeping

Aerial Ability

On Ball

Off Ball

Fouls

Shot Stopping

One-on-Ones

Distribution

Decision Making

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

27 of 37

Rating Players Based on a Specific Game Model

INSIGHT DELIVERY TO FOOTBALL STAKEHOLDERS

27

10

11

8

3

2

5

4

6

7

9

1

HIGH PRESSING

LINK UP PLAY

ATTACKING FIRST POST FOR CROSSES

RUNS IN BEHIND

STRETCH LINE OR FALSE 9

Striker

OFFENSIVE

GOOD FOOTBALL

FROM THE BACK

ORGANISED, DYNAMIC

3-2-2-3

In Possession

NUMBERS REPRESENT PLAYER PROFILES

28 of 37

28

DATA ENGINEERING AND INFRASTRUCTURE

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

29 of 37

Licensed and Public Data Providers Accessible to the Club

CURRENT STATE AND RECENT PROGRESS

29

Event

Data

Scouting and Financial Data

Tracking

Data

Sports Science and Physical Data

Additional Data

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

30 of 37

AWS Cloud-Based Data Infrastructure and Tooling

DATA ENGINEERING AND INFRASTRUCTURE

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

31 of 37

Building the Player 360 Database

INSIGHT DELIVERY TO FOOTBALL STAKEHOLDERS

31

Event data

(40+ competitions)

Optical Tracking data

(Premier League)

Physical data

(Leicester City only)

Additional Data

(TransferMarkt, WhoScored Ratings, SmarterScout metrics, etc.)

Player 360 Reference Table

table of common identifiers containing each player's key details and unique ID across the systems, allowing data to be matched

Well-being and sports science data

(Leicester City only)

32 of 37

32

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

OPEN SOURCE CONTRIBUTIONS

Summary

33 of 37

Overview of Open Source Projects

OPEN SOURCE. CONTRIBUTIONS

33

Recruitment analysis of ‘ball-playing’ centre backs

Scenario for a data-driven recruitment analysis, to determine ”the next Gerard Piqué”, for a hypothetical, newly-promoted club in the ‘Big 5’ European leagues. This analytical piece features the use of Machine Learning techniques as part of the player selection process, including K-Means clustering and Principal Component Analysis.

Opposition analysis of a Premier League team through the application of Event and Tracking data

Application of Opta Event data and Second Spectrum Tracking data for two of Crystal Palace’s matches in the 21/22 season against Brighton & Hove Albion and Leicester City, as part of an opposition analysis report for a Premier League club.

Physical data analysis of player GPS training data

Working applications of player training data for the purposes of performance analysis in a club setting, with an output of player reports to measure speed, time, distance covered, acceleration, and number of sprints that players conducted in each of the training sessions they participated in.

Post-tournament analysis of the England men’s team at EURO 2020 through the application of StatsBomb Event and OBV Data

Short analytical piece that applies a data-driven approach in an attempt to answer the question ‘England at EURO 2020: How Good Were They Actually?’.

Tableau Football Intelligence reporting tools

Creation of analysis tools for analysing player performance, pre- and post-match tactics, and recruitment, enabling the democratisation of data to all levels of a footballing organisation or club. Data used features StatsBomb Event data for the FA WSL for the 19/20 and 20/21 seasons.

Applications of Tracking data

Engineering and analysis of Signality, Metrica Sports, and Stats Perform Tracking data. Analysis includes analysing passages of play, determining physical performance, the implementation a basic Pitch Control models, and applying Expected Possession Value (EPV) frameworks.

Expected Goals (xG) modeling

Training an Expected Goals (xG) model using Event data, subsequently applied to a separate match dataset of Event and Tracking data, to analyse the performance of the teams in question. Models created using Logistic Regression, Random Forests, and Gradient Boost Decision Trees (GBDT) algorithms including XGBoost and CatBoost.

1

2

3

4

5

6

7

These projects are all discussed in further detail in my Data Science Portfolio: docs.google.com/presentation/d/1EQ4meB5Y2TGuIn3K31OdkvoI-8QlPW2MhZE8UEhZ38c.

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

34 of 37

34

SUMMARY

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

35 of 37

Insights this Football Data Portfolio Aims to Provide

SUMMARY

35

Data insights experience – this portfolio aims to have provided cross-section of my five years of data insights experience, developed from elite football environments and the commercial departments of top-flight football club, as well as best-in-class industries.

Diverse skill-set and problem-solving approach – breakdowns are provided for tackling common industry challenges with working examples, covering the concepts of data processing, modelling, management, visualisation, and strategic planning.

Measurable impact – demonstrating the quantifiable outcomes for football analytics data and insights.

1

2

3

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

36 of 37

  • LinkedIn: linkedin.com/in/eddwebster

  • Twitter: @eddwebster

  • Email: edd.j.webster@gmail.com

For More Information…

SUMMARY

36

Introduction

Areas of Impact

Football Intelligence

Insight Delivery to Stakeholders

Player Ratings Framework

Data Engineering and Infrastructure

Open Source Contributions

Summary

37 of 37