xGenius, page 2
Chapter 10 investigates the world of set pieces. Dead-ball situations are the low-hanging fruit which analytical clubs have grabbed with both hands. Mastering the art of set pieces can improve your goal difference by 20 goals each season. It is often the difference between winning and losing.
We have data on hundreds of thousands of shots, each one a unique experiment into how football works.
Chapter 11 takes a closer look at the dynamics of ‘finishing ability’. Do players actually possess a God-given talent at finding the back of the net, or do some simply benefit from the favour of Lady Luck?
Chapter 12 examines a deeper analytical toolkit which we can use to illuminate the beautiful game. Expected Goals has become the poster boy for football analysis, but there are several other metrics worth exploring if we’re going to develop a full understanding of the science of winning matches.
Chapter 13 ventures into the dressing room. In a game so increasingly focused on individual talent, the output of teams is often ignored. Can we quantify the impact of team chemistry? And can we measure how well all 11 players on a field are synchronised in their movements, their ability to create space for one another? This is perhaps where the science of football is at its most advanced.
Chapter 14, the final section, reflects on how the tidal wave of xG has swept over the shores of football in recent years, as well as looking towards the future of sports analytics. The same philosophy which has brought football teams success is now being applied to other games. Golf, basketball, cricket, hockey, tennis and many other pastimes have adopted an xG way of thinking which is helping coaches and players win more often.
A decade ago, this book couldn’t have been written. Although many of the questions posed – those of finishing ability, playing style and transfer strategy – have been around for decades, we’ve only recently unearthed the data required to provide the answers. Chance creation has always been at the centre of football. Journalists wrote match reports centred around the key goalscoring opportunities at either end of the pitch long before the advent of three-minute-long YouTube highlight packages. But the invisible hand of xG has only recently revealed itself in the flesh. Now we are able to count and quantify the ability of teams to create dangerous situations. We have data on hundreds of thousands of shots, each one a unique experiment into how football works. This has completely transformed the landscape in terms of how we evaluate and think about footballing performance. Other books on this subject have told the story of human beings who use football statistics. This book aims to make the data itself the central character. Hopefully xGenius offers the reader a new perspective, like wearing x-ray glasses to reveal the hidden structures and forces behind the chaotic and messy reality of the sport.
As well as answering some of the puzzles we’ve been trying to solve for generations – ‘how good actually is my team?’, ‘where do goals come from?’, ‘what shots are most valuable?’ – the dawn of the xG era has also introduced new ones – ‘does finishing ability exist?’, ‘how much impact does game state have?’, ‘what balance should clubs strike between data-driven methods and traditional ways of thinking?’
Football is ‘the infinite game.’ No one will ever be able to fully understand all its nuances, and no one will ever be able to ‘master it’ or ‘complete it.’ All we can do is get as close to the truth as possible. Think of each football match as an image. Looking at it simply through the lens of the final scoreline means the image is incredibly blurry. When you begin to layer in conventional stats – possession, shots, shots on target – you begin to improve the focus. Introducing the sort of advanced metrics we’ll study in this book – Expected Goals, Field Tilt, Expected Possession Value – will provide even more clarity. The aim of football analytics is to make the image as high-resolution as possible. This book explores the interplay between analysis, tactics, and decision making. It seeks to put the sport under the microscope with the aim of getting closer to the ultimate truth of what makes players, managers, and teams successful. What, ultimately, wins football matches.
1
The Model Football Club
The Teams that Changed the Game
‘Innovation, not increased funding, can be the only route to success for clubs such as ours’
Matthew Benham, Brentford FC owner
Graham Potter puffed out his cheeks and tilted his head to the sky as Christian Benteke wheeled away in celebration. Brighton & Hove Albion had just conceded a 95th-minute goal to go 2-1 down against their bitterest rivals. Potter was having one of those nights that every football fan has endured at some point. His team had dominated the game, taking 25 shots to Palace’s three. They had completed 13 passes within 20 yards of their opponent’s goal, conceding zero of such passes. Brighton had created far better scoring opportunities, accumulating 2.06(xG) to Palace’s 0.24(xG). If the exact same match was played out one hundred times, Brighton would have won on 93 occasions and lost just once.1
Brighton travelled to West Bromwich Albion five days later. Kyle Bartley headed the hosts in front after 10 minutes with an effort that could generously be described as a half chance. Brighton laid siege to their hosts over the course of the next 80 minutes, missing two penalties and a host of other big chances. The match finished in a 1-0 defeat, despite an Expected Goals scoreline that read 3.14(xG) to 1.13(xG) in favour of Graham Potter’s men. For the second consecutive match, the chance of Brighton losing based on the game’s scoring opportunities rounded to 1 per cent. For the second consecutive match, that 1 per cent likelihood took place.
A week later, Brighton lost their third match in a row. A tightly contested game once again saw the gods rule against them as they fell to a 2-1 defeat to Leicester City. The loss left them 17th in the Premier League table, clear of the relegation zone only by virtue of a superior goal difference to Fulham, albeit having played one game fewer. At this point in the season, early March 2021, Brighton had played 27 matches. They’d scored 16 goals fewer than expected and conceded five goals more than expected, according to xG. They’d accumulated 26 points, but leading Expected Goals models indicated they should have amassed roughly 46 points based on the quality of their performances throughout the campaign. Brighton would have occupied fourth place if the table were ranked according to Expected Points, behind only Manchester City, Chelsea, and Liverpool.
Brighton’s extraordinary run of form throughout the 2020/21 season can be described using a host of adjectives. Unlucky. Implausible. Brutal. But overall, it was funny. It was funny in the ridiculousness of it, in the absurdity of it. We often make exaggerated claims about sport having the ability to write scripts that no human could conjure up, but surely even the most imaginatively cruel author couldn’t invoke the tale of Brighton’s woes that campaign. They were playing teams off the park and creating an enormous number of chances but ending up on the wrong side of the result time and time again.
This tragicomedy playing out on the Premier League stage caught the imagination of social media. One Twitter account (as it was then known) was perfectly placed to tell the narrative of Brighton’s tale of anguish. ‘The xG Philosophy’ (handle of @xGPhilosophy) had started posting post-match Expected Goals scorelines when football returned from the Covid break. The first significant Brighton result of note came in late September 2020, when they lost 3-2 to Manchester United despite creating 3.03(xG) to United’s 1.91(xG). (This game also famously saw United awarded a penalty via VAR after the full-time whistle had blown, which Bruno Fernandes duly converted to win the match in the 100th minute). From there, the cult perception of Brighton as ‘the xG team’ gained increasing momentum with each undeserved defeat. Memes were shared which jeered Brighton’s demise and gained huge traction across ‘football Twitter’.
The perception of Brighton as ‘the xG team’ gained increasing momentum with each undeserved defeat.
This saga had wider consequences than simply the mockery of a south-coast football club. There was still an inherent distrust of xG from the wider football community at this point in the evolution of football analytics. The comments under every @xGPhilosophy post at the beginning of the 2020/21 season skewed more towards rejection than they did acceptance, but Brighton’s performances managed to turn that tide. The non-believers were watching Brighton’s matches, seeing their incredible dominance, then having visibility of the xG stats after the game. The data was validating the eye test. Brighton’s results weren’t aligning with their performances and the xG deniers began to understand what the metric was all about.
Engagement began to snowball. Starting from an initial base of 8,000 followers in June 2020, @xGPhilosophy grew to over 200,000 in the space of a year. The new followership included notable figures within the world of punditry such as Jamie Carragher, David Jones, Micah Richards, Michael Owen, and Xabi Alonso. The former two of these began integrating xG into Monday Night Football, Sky Sports’ leading football analysis show. The craze even sparked the attention of Brighton’s official Twitter account, which took notice of the cult following their team had built in this space and began self-deprecatingly replying to @xGPhilosophy’s posts on a regular basis.
Brighton did end up avoiding the cold grasp of relegation that season, finishing in 16th position on 41 points. Based on their xG performances, leading models estimated they should have finished on 62 points, the fifth-best Expected Points total of any team in the league. Their underperformance versus expectation was the most severe of any team since xG data has become available, and the most severe we will likely see for quite some time. The cruel irony of this story is that Brighton are pioneers in the global football analytics revolution. They were, in fact, one of the first teams to harness the power of xG.
Climbing the Pyramid
Rewind back to before the turn of the millennium: Brighton were a club in dire straits. Having lost the Second Division play-off final in 1991, the club entered free fall. Two relegations later, The Seagulls found themselves needing a win against Hereford United on the final day of the 1996–1997 season to avoid relegation to non-league. A late equaliser from Robbie Reinelt ensured Brighton retained their league status by the tightest of margins. The turn of the millennium saw Brighton forced to move to Withdean Stadium, a converted local athletics track which was owned by the council. The capacity of the ground was a measly 6,000 initially, increasing to 8,500 after a couple of years. It was voted the fourth-worst football stadium in the UK by The Observer in 2004. The temporary nature of the ground was obvious. It was primarily used for athletics, with a single permanent stand along the north side and the remainder of the stands assembled from scaffolding. Changing and hospitality facilities were provided by portable cabins placed haphazardly around the site. The ground summed up Brighton’s status as a football club in the early 21st century. This unhappy stage of their history was spent bouncing around the lower echelons of the Football League.
But one day, everything changed.
Tony Bloom allegedly placed his first bet at the age of eight, when he would use his pocket money to play the fruit machines at the nearby arcade. Bloom’s passion for gambling continued well into his teenage years. At the age of 15, he regularly snuck to the nearest city to make use of a fake ID for betting purposes. He went on to study Statistics and Mathematics at The University of Manchester, before securing a job as an accountant upon graduation. His career path led him from the trading room floors of the city to those of the bookmaker Victor Chandler (which would later become BetVictor). Here, Bloom learned his craft, acquiring a profound understanding of the betting markets while also tinkering with his own model for predicting the outcome of football matches. He eventually went out on his own, setting up several online gambling sites during the early 2000s and occasionally taking time out to play poker against the world’s elite. Bloom’s first major win came in 2004 when, at the age of 33, he won the Australasian Poker Championship in Melbourne, collecting the first prize of £180,000. By 2008, his live tournament winnings exceeded £1,200,000. His cold-blooded nature, expressionless stare and cool decision-making bestowed upon him a nickname: ‘The Lizard’.
Bloom drew inspiration from this nickname when founding a betting consultancy, Starlizard, in 2006. The company’s success was based on Bloom’s top-secret model for predicting match outcomes. In simple terms, they were first to the xG party and managed to drink all the free booze before the rest of the guests turned up. Starlizard used, and continues to use, Expected Goals data to form accurate gauges of a team’s ability. Their xG data allows them to identify undervalued selections in the betting markets and has enabled them to make hundreds of millions of pounds in profit. By 2009, Bloom had acquired enough capital to swoop in and rescue his childhood football club, the struggling Brighton & Hove Albion. Bloom applied the same statistical analysis that brought him success in the world of betting to his running of Brighton. Expected Goals data informed Brighton’s decision-making across several verticals; from recruitment, to style of play, to opposition scouting.
When Bloom took over in May 2009, Brighton were about to embark on their fourth consecutive season in League One, the third tier of the English pyramid. A couple of years later, in their final season at Withdean Stadium, the club secured promotion to the Championship. They moved to a brand-new ground called Falmer Stadium (later renamed the Amex for sponsorship purposes) and duly changed their crest to a design similar to that used from the 1970s to the 1990s, reflecting the club’s return home having not had a stadium of their own since 1997. The rise did not stop there. Shrewd recruitment, a restructuring of the management staff and the channelling of statistical analysis allowed the club to defy the odds and achieve promotion to the Premier League for the first time in 2017, just two decades after they were on the brink of elimination from the Football League (although they played in the top flight from 1979 to 1983, when it was known as the First Division). Since then, Brighton have been a mainstay in the richest division in the world, despite possessing a budget which is dwarfed by many of their competitors. Their infamous 2020/21 xG car crash of a campaign was followed up with their highest ever league finish in 2022, ninth in the Premier League. In 2022/23 they managed to achieve a European spot by finishing sixth, despite possessing the third-lowest wage budget in the league.
Brighton aren’t the only club operating in the analytical half-space. Although all Premier League teams now possess some form of analysis department, three English clubs stand apart from the rest in their wholesale adoption of analytics. Of the triumvirate, Brentford FC are the most similar to Brighton in their structure and approach. In fact, the two teams are football’s equivalent of identical twins. Both clubs are run by a fan who made their millions by applying a rigorous statistical approach to the betting markets. They’ve both adopted an xG-driven approach to running their clubs and have structured their organisations to tailor for a ‘data first’ philosophy. Both clubs have enjoyed an incredible rise up the Football League pyramid as a result of outthinking, rather than outspending, the opposition (see Figure 1.1).
Figure 1.1: Brentford and Brighton League Finishing Positions, English Football League 2006–2023
For Tony Bloom, read Matthew Benham. For Starlizard, read Smartodds. Bloom and Benham were initially colleagues at an online bookmaker that Bloom founded called Premier Bet. Benham worked under Bloom, but the pair had an acrimonious falling out. After leaving Premier Bet, Benham founded Smartodds in 2004. Two years later, Starlizard came to be. The offices of the two companies are just down the road from one another, in Kentish Town and Camden respectively. North London has a strong claim of being the Wall Street of professional football betting.
Matthew Benham’s football club has followed a remarkably similar rise as Tony Bloom’s. Brentford were a mid-table League One team when Benham took the reins in 2012. Two years later, they were promoted to the Championship for the first time in 20 years. Despite possessing the fourth-smallest playing budget in the 24-team division and being heavily tipped for immediate relegation, the club finished fifth in their first season in the second tier. After a few years of consolidation at that level, Brentford reached the top division for the first time since 1947. Their first season in the top flight saw them finish 13th in the table, despite being the poorest club in the league. The team then defied all logic by achieving a top half finish in 2022/23. Brentford have been transformed from a mid-table third division team into a comfortable Premier League club in less than a decade. An embracing of analytics appears to carry an anti-gravitational, upwards force that shoots your team to the top of the league pyramid.
Wage expenditure has long since been recognised as the strongest indicator of a team’s performance. The deeper the owner’s pockets, the better players their team can attract, meaning the higher up the league table they finish. In 2022/23, Brighton had the second-lowest annual wage bill of the 20 Premier League clubs. Despite this, their performances on the pitch led to a sixth-place finish and a Europa League spot. Their xG performances were even better. According to Expected Points, Brighton should have finished in the top four positions and earned a Champions League spot. Although possessing the wage budget of a relegation-threatened team, they were actually the fourth-best team in the league that season according to Expected Points.2
Figure 1.2: Expected Points per Game vs Annual Wages3, Premier League 2022/23
Brentford were equally as impressive, with an annual wage bill of just £34 million. According to every rule of football logic, the Bees should have finished in last place. Instead, they narrowly missed out on a Europa Conference League spot on the final day of the season. They beat treble-winning Manchester City twice, the only team to do the double over Pep Guardiola’s side that season. They also defeated Manchester United, Liverpool, Tottenham Hotspur, and Chelsea, meaning they had the best record against the ‘big 6’ teams of any side that season. Expected Points revealed them as the seventh-best team that campaign and worthy of securing European football.
