Football Hackers brilliantly charts the journey of football from being perceived as a ‘simple, emotionally-charged game’ to how increasingly complex it has become, as we look closer and closer.
At the beginning of the book, Biermann dispels some of the common footballing biases, such as the outcome bias, which refers to our tendency to construe judgements based on results. If the outcome is positive, we assume that the manager’s plans were right. But If the result is negative, we are quick to believe that his tactics must have been bad. Thinking that way disregards the role of luck in football and the fact that there’s quite often a gap between performances and results. This is where the likes of xG (Expected Goals) and other metrics come in, as they try to eliminate luck and quantify performance.
Biermann doesn’t make lofty claims, the way James Tippet does in The Football Code, that xG is the one-stop solution for the world’s problems. He acknowledges that data analytics in football is still evolving and is not quite the finished product as it is in American sports or Formula 1, which incidentally is the most technological sport in the world, with dozens of sensors producing a staggering 10TB of data for every race.
The book explains xG with a popular example, leveraging Colin Trainor’s groundbreaking analysis of Jurgen Klopp’s last season with Borussia Dortmund, when one of Europe’s best teams inexplicably got so bad as to get mired in a relegation battle. Trainor writes that ‘the use of analytics can help us begin to make assessments about whether certain results have arisen from great skill or were possibly due to a combination of fortuitous circumstances’.
In Dortmund’s case, he found the numbers pointed towards a random sequence of bad luck. They had created enough good goalscoring opportunities to finish the year with 25 goals instead of 18. At the other end of the pitch, they should have conceded 17 instead of 26. Taking both numbers together, Dortmund should have had a goal difference of +8 rather than -8. Extrapolating Expected Points (xP) from those figures, Trainor calculated that Dortmund should have amassed a tally of 30 as opposed to the meager 15 at the winter break and should have sat in fourth place when Klopp was sacked.
Another good example cited in the book is Newcastle United over 2 contrasting seasons. In 2011-12, the Magpies, under Alan Pardew, finished in fifth spot against all odds. After receiving plaudits for a great season, Newcastle crashed down to 16th spot in the following campaign and barely survived relegation. In the first season, there had been much talk about the side’s attractive play and Pardew’s courage to play youngsters. The next season, their style was considered too gung-ho and that there was a lack of experience in the side. This is another classic example of outcome bias, the tendency to judge efforts on their results and the blind assumptions that good results are underpinned by good performances and that bad results are the consequence of bad performances. Newcastle’s case was especially bizarre, since both assumptions were wrong. According to their xG and the resulting distribution of points, their performances in both seasons hardly differed. In both campaigns, Newcastle had created and conceded a similar amount and quality of goalscoring opportunities. So, they were really being applauded for being lucky and blamed for being unlucky.
The book is lucid in its praise of the one manager who manages to outwit the xG gods time and again, Borussia Dortmund’s Swiss coach Lucien Favre, who has enjoyed hugely successful stints with Monchengladbach and Nice as well. Favre’s teams dramatically overshoot Expected Goals, both in goals scored and goals conceded. In the 2016-17 Ligue 1 season, his Nice side should have had a goal difference of -10 according to Expected Goals. But they had +27, a swing of 37 goals! Favre might simply have been the luckiest manager in Europe that year. But this wasn’t a one-off. In 2011-12 and each of the two subsequent seasons, he beat the Expected Goals model, that too with a good margin.
xG provides pointers for the gap between performance and yield, as already explained. But if Favre’s teams were consistently more successful than anticipated, there must be a blind spot in the model and the entire football analytics community was eager to unravel the mystery. The most extensive attempts were made by Michael Caley, a pioneer in xG models and a Spurs fan, and India’s leading football analyst, Ashwin Raman.
The value xG calculates for a shot is a mathematical approximation. So, the model does have a blind spot as it doesn’t fully take into account the opposition. It makes for a huge difference, of course, if a striker is through on goal with all the time in the world or if he’s being challenged by a centre-back who’s been instructed to use his stronger left foot to tackle by his coach Lucien Favre.
Ashwin picked up on another interesting detail. Favre’s men took nearly 60 percent of their shots when only two or fewer opposition players (goalkeeper excluded) were blocking their path to goal. There was a mirror image in defence. Opponents were forced into shots with high defensive pressures and a high number of players between shooter and goal. Because of that, opponents didn’t really benefit from taking shots from high probability positions. Favre’s game adhered to a simple idea: it ensured that his team took good shots and his opponents bad ones.
Football’s transformation into a game of numbers started with the founding of ProZone in 1995. The company’s first client in the Premier League was Manchester United, who were having the best season in their history, winning the treble of Champions League, Premier League and FA Cup. Within 18 months, a host of other PL sides also employed ProZone’s services, chief among them Sam Allardyce’s Bolton Wanderers. Inspired by American sports, ‘Big Sam’ and his analysts developed a data-based playing style that he hoped would help Wanderers survive in the Premier League following their repromotion in 2001. They knew that dead balls accounted for up to a third of all goals and were more dangerous if the deliveries were played towards the opposition goal (i.e. in-swingers). Bolton practiced freekicks and corners extensively, and employed throw-ins, too, to create havoc in the box. Based on other similar principles, they made the unfashionable club from Greater Manchester fantastically successful in relative terms. Between 2003 and 2007, they regularly finished in the top eight, twice qualifying for the UEFA Cup, albeit not playing an attractive playing style.
No book on football analytics is complete without a mention of Matthew Benham, the closest thing to ‘Moneyball’ in football. It’s well documented how Benham, the owner of English Championship side Brentford FC and Danish club FC Midtjylland, uses pioneering analytical methods to recruit players, allowing his teams to consistently sign undervalued prospects. With his acquisition of Midtjylland, Benham found the perfect test lab for his novel methods. The collaboration with Benham’s SmartOdds enabled the club to evaluate its players using KPI (Key Performance Indicators). In addition, they were able to analyze shooting positions that afford the highest-scoring probability, with a view to directing the team towards moving the ball into those areas. It was Expected Goals, reverse-engineered: if one shooting position was doubly promising than another one, it made sense to figure out how to get there.
Biermann also talks about Benham’s former friend, Tony Bloom, who founded Starlizard, a hugely successful betting company. Like Benham, he also bought his boyhood club, Brighton & Hove Albion, before building a new stadium for the club and leading them to the Premier League in 2017.
Football Hackers also covers some of the lesser-known ‘Billy Beanes’ such as Chris Anderson, a US academic, who started pursuing football analytics after reading Moneyball. After futile attempts to convince US-investors to buy a Premier League side, he went on to become the Managing Director of Coventry City, then in the third division. Unfortunately for Anderson, any fantasies of becoming Billy Beane did not survive the first brush with reality and he left Coventry and went back to academics after just over a year in the job.
Football Hackers covers several other football innovators and even some non-xG metrics such as Packing stats, all narrated through well-constructed stories and real-life anecdotes. The book is heavily recommended for modern fans of the beautiful game, who want to join football’s data revolution and understand how the arrival of advanced metrics and detailed analysis is already reshaping football.