How the spreadsheet-wielding geeks are taking over football

The statistical revolution comes to the pitch.

You are invited to read this free preview of the upcoming New Statesman, out today. To purchase the full magazine - with our signature mix of opinion, longreads and arts coverage, plus columns by Helen Lewis on Japan, Will Self on Claire Balding, and a major report from Turkey by Helena Drysdale - please visit our subscription page.

 

 

The Numbers Game: Why Everything You Know About Football Is Wrong
Chris Anderson and David Sally
Viking, £12.99, 384pp

Almost exactly a decade ago, the American writer Michael Lewis published a book called Moneyball. It told the story of Billy Beane, the general manager of an unfashionable baseball team, the Oakland A’s, who was using new statistics to evaluate baseball players and strategies. From this unpromising material, Lewis crafted a bestseller that has sold more than a million copies.

Books hardly ever change anything but this one did. Moneyball changed baseball and almost all ball games from basketball to cricket but it also affected worlds beyond sport. Ken Mehlman, Republican campaign manager in the US presidential election of 2004, instructed his staff to read it, because he realised that it wasn’t just a sports book. It was also a perfect case study of how crunching numbers can give you an edge. That made it a book for our era of “Big Data”, in which the amount of data on earth more than doubles every two years and the only mystery is how to use all these confusing numbers.

Football, or “soccer” was always the most hidebound sport and it held out longest against the numbers revolution. But now, as Chris Anderson and David Sally write in their engaging and stimulating book The Numbers Game, “The datafication of life has started to infiltrate football.” That is quite a change. In football, people always did what they did because they had always done it that way. Clubs were historically run by autocratic managers who had left school at 16 to become players and didn’t hold with book-learning. However, unseen by most fans, something profound is happening inside the sport.

Like Moneyball, The Numbers Game reaches us from the US, which is rapidly becoming a soccer society. Anderson, originally from Germany, played semi-professional soccer before becoming a professor of government at Cornell. His neighbour Sally, once a baseball pitcher at Harvard, is a behavioural economist at the Tuck School of Business at Dartmouth. Watching soccer on television together, they grew interested in the game’s relative lack of numbers and analytics.

That absence had struck the first pioneer of numbers in football, Wing Commander Charles Reep. Not a fighter pilot but an accountant in the RAF’s Bomber Command, Reep made what is probably the first known attempt to log “match data”.

In 1950, at a Swindon Town game, he logged 147 attacks by Swindon in the second half. Extrapolating from this small sample, Reep calculated that 99.29 per cent of attacks in football failed. He continued to offer his services as an analyst to clubs into his late nineties but, as Anderson and Sally show, he was on a wild goose chase. Reep assumed that there was only one correct way to play football and, naturally, he thought he had found it. Boot the ball long, said the wing commander – put it near the opposition’s goal and you will win.

In reality, Anderson and Sally write, “There is no winning formula. There is no right answer to football.” Different strokes suit different teams. To quote the great Liverpool manager Bob Paisley: “It’s not about the long ball or the short ball; it’s about the right ball.” (The best managers of the past, including Paisley, intuited many of the findings now emerging from the numbers.)

In the mid-1990s, the spread of computers reignited the data revolution. Companies such as Opta and Prozone began collecting stats on football matches. Suddenly, clubs knew how many passes each player had completed, how many tackles he had made and how many kilometres he had run.

As soon as data becomes available in any industry, some people will use it – but they often use it wrongly. As the American baseball analyst-turned-master psephologist Nate Silver says of the new Big Data: “Most of the data is just noise, as most of the universe is filled with empty space.” Alex Ferguson, Manchester United’s manager, discovered this after he sold his defender Jaap Stam in 2001 because Stam’s number of tackles was decreasing. Ferguson thought Stam was in decline. Stam went on to play several more years for big clubs.

It turned out that tackles were a poor measure of a defender’s worth: they were just noise. We now know that great defenders such as the Italian Paolo Maldini barely tackle. Maldini stopped attacks from happening by positioning himself to close holes. Yet, as Anderson and Sally point out, that kind of negative event – the attack that doesn’t happen, the dog that doesn’t bark – is often hard to spot in match data. Football statistics tend to focus on things that do happen and, above all, on goals that do get scored.

As the data revolution progresses, more and more clubs are finding clever ways to use numbers. Each season, the number of sceptics declines – in part because many people in the game have now read Moneyball or at least seen the 2011 Hollywood movie starring Brad Pitt as Beane.

Not all the traditionalists are going quietly into the night. Some are now scheming to defend their turf against spreadsheet-wielding geeks. But others are learning not to believe their own eyes. As Beane told me: “The idea that I trust my eyes more than the stats – I don’t buy that, because I’ve seen magicians pull rabbits out of hats and I just know that rabbit’s not in there.”

The data revolution keeps stumbling on new truths. At Manchester City, for instance, the analysts finally persuaded the club’s then manager, Roberto Mancini, that the most dangerous corner kick is the inswinger, the ball that swings towards goal. Mancini had long argued (strictly from intuition) that outswingers were best. Eventually he capitulated and, in the 2011-2012 season, when City won the English title, they scored 15 goals from corners, the most in the Premier League. The decisive goal, Vincent Kompany’s header against Manchester United, came from an in swinging corner.

The most powerful figure in English football remains the manager and the statistical revolution has progressed fastest at clubs where the manager believes in data. Probably the leaders in this field in England today are Arsenal’s Arsène Wenger (an economics graduate and gifted mathematician), West Ham’s Sam Allardyce (not a gifted mathematician) and Manchester United’s incoming manager, David Moyes.

In March, I visited Moyes’s then club, Everton, and one of his data analysts told me, “In terms of managers, he is probably as into it [data] as any.” Moyes would often march into the analysts’ offices firing out questions: how efficient were Everton’s next opponents at scoring from crosses? What types of passes did their midfielders make? In which areas of the field did Tottenham’s superstar Gareth Bale usually receive the ball?

For managers such as Moyes, data isn’t everything. It is one tool among many. It gives you an edge and, since you could employ perhaps 30 statisticians for the £1.5m that the average player in the Premier League earns, it’s an edge you can afford. Still, as Anderson and Sally caution: “The data cannot do the manager’s job.” Interpreting data is an art more than a science.

In 2004, the data told Wenger that an unknown French teenager, Mathieu Flamini, was running an astonishing 14 kilometres a game. By itself, that number wasn’t enough. Did Flamini run in the right direction? Wenger went to watch him, decided he did and signed him for peanuts.

Even more cheaply than hiring another statistician, a cunning manager could pop into a bookshop and splash £12.99 on The Numbers Game. The book contains several fascinating examples of statistics that could help club chairmen, managers or fans. Perhaps the book’s most remarkable finding is that football is a “weakest-link game” – although it’s nice to have great players in your team, it’s more important not to have rubbish players. Games are typically decided not by the Wayne Rooneys but by oafs such as Zurab Khizanishvili, a defender whose blunders in a play-off in 2011 arguably cost Reading promotion to the Premier League.

Anderson and Sally crunch some of the new data on individual players to estimate that upgrading your weakest link typically improves your team more than buying a new superstar would. Despite this, managers, being human and wanting to please fans and journalists, usually prefer the superstar.

The book overturns several other tenets of football thinking. For instance, the old saying that you’re most likely to concede a goal straight after scoring turns out to be nonsense. According to the stats, that’s when you’re least likely to concede.

The numbers also show the outsize role of chance in football. In one study of 43,000 matches, the underdog won 45.2 per cent. Favourites win much less often in football than in other ball games.

That is chiefly because goals in football are so scarce: you can attack all match but if the opposition nicks one lucky goal, you can lose. Then the media and fans provide a post hoc rationalisation for your defeat, even though it was dumb luck.

The Numbers Game also shows that sacking the manager – football’s equivalent of the human sacrifice – is usually pointless. Typically, the manager is sacked when the team hits its lowest point. Yet any statistician can predict what will happen after you hit your lowest point: performance will improve, because of the statistical phenomenon known as regression to the mean. Anderson and Sally explain: “An extraordinary period of poor performance is just that: extraordinary. It will auto-correct as players return from injury, shots stop hitting the post or fortune shines her light on you once more.”

Sunderland briefly improved this spring under their new manager, Paolo Di Canio, not because fascism works but because of regression to the mean.

There is an excellent final chapter predicting how football’s data revolution will progress. The authors forecast, for instance, that the historic undervaluation of goalkeepers and defenders – who command lower average salaries and transfer fees than strikers – is likely to end.

That’s because stats show that keeping a “clean sheet” helps a team more than scoring lots of goals does. And, as data evolves, we will find ways to value the almost invisible contributions that defenders such as Maldini make. Anderson and Sally believe that football data will increasingly focus on the geometry of the game off the ball – which is crucial, as the average player has the ball for only 53 seconds a game.

They also predict that the biggest innovations will come from poorer clubs, football’s equivalents of the Oakland A’s: “The strong do not need to innovate; it is the weak who must adapt or die.” Rich clubs such as Chelsea can succeed simply by buying great players. As the authors admit: “Analytics will help you win, but so will money.”

Moneyball was the Communist Manifesto of the data revolution, in sport and beyond. The Numbers Game isn’t as groundbreaking as its authors proclaim. Its subtitle – “Why Everything You Know About Football Is Wrong” – is an unnecessary overstatement. Nonetheless, the book is a valuable addition to the scarce literature at a time when pioneers inside football are only just starting to work out which stats matter, while people outside the game still scarcely know that anything is changing.

The Numbers Game is energetically and cleanly written and is free of academic jargon, though it is occasionally guilty of faux-poetic overwriting: “Each side possesses a light side, seeking the goal, and a dark side, hoping to divert it. And at the centre of that collision between the positive and the negative, the yin and the yang, is the ball” – and so on.

The authors have done their homework and I have only one sad correction to make: Nick Broad isn’t a performance scientist with Paris Saint-Germain any more. He was killed in a car crash in January, aged 38.

Simon Kuper is co-author of “Soccernomics” (HarperSport, £8.99) and a columnist with the Financial Times

An aerial view of the Hackney Marshes football pitches in London. Photograph: Getty Images

This article first appeared in the 10 June 2013 issue of the New Statesman, G0

Photo: Justin Tallis/Getty Images
Show Hide image

What does our latest poll mean for the Labour leadership race?

Jeremy Corbyn is ahead among councillors - and looks ever more certain to become Labour's next leader. 

This morning the Labour History Research Unit at Anglia Ruskin University released its last set of polling data of Labour councillors in marginal constituencies’ prior to the election of the new leader.

It’s certainly a limited enough snapshot but in broad terms the data suggests four things. Firstly, that Jeremy Corbyn will win the leadership. Perhaps no great shock there at this point. But Corbyn’s slight lead in our poll of only two points or under above Yvette Cooper and Andy Burnham masks the fact that he has picked up over 11 per cent of councillors since June - whilst all other candidates have lost support here. Given his reputation as a centraliser, it is remarkable that Corbyn is also neck and neck with Andy Burnham as the candidate councillors believe ‘would be best for local government.’ If he’s just about won over this tough crowd it may indeed be game over.

Secondly, the £3 registered supporter experiment is viewed as a damaging one by many within the party. With almost six in ten councillors thinking it should be ‘scrapped ahead of any future contest’ compared to just over one in four seeing it as a positive, there may well be clamour to reform this model going forward. Whether Corbyn will want to challenge the legitimacy of a reasonable proportion of his backers is one thing, but he would likely have some support in doing so if others were to press the issue.

Thirdly, on whatever mandate Corbyn is elected the good news for him is that key councillors clearly back Corbynomics. His plan to create a regulated and publicly-run service to deliver energy supplies is backed by 78 per cent of councillors who either “strongly agree” or “agree” with the policy, while 77 per cent support nationalising the railway network as soon as practicable. Introducing a 50p top rate of income tax is backed by 79 per cent of councillors, while 73 per cent agree with a “mansion tax” on homes worth over £2million. Most of those individually poll well amongst the electorate, though the 75 per cent of councillors who think scrapping tuition fees would aid the Labour vote in their constituency are out of kilter with the only one in six members of the general public who support that measure.

But lastly, perhaps most crucially, the rub is that less than two in ten councillors surveyed think Jeremy Corbyn will win the 2020 General Election. Even amongst councillors pledging to vote for Corbyn that figure tops out at six in ten.

Our data aside, Corbyn’s medium term challenge will clearly be enormous, as they would be for any new leader. For one, Labour’s current core vote just doesn’t turnout in enough numbers – not only in terms of voting for Labour, but at all. In 2010 and 2015 Labour’s most successful demographics were the semi-/low skilled working class (40 per cent to 31 per cent over the Tories in 2010, 41 per cent to 27 per cent in 2015) and ethnic minorities (60 per cent to 16 per cent in 2010, 65 per cent to 23 per cent in 2015). Turnout for both these groups is at least one in ten less than the national average, and barely bobs over one voter in two generally.  

Instead, in 2015 the most likely people to vote were men over the age of 55 (79 per cent), the middle class (75 per cent), or property owners (77 per cent). And so Jon Cruddas’ reviews’ conclusion that Labour has fallen behind on the average Prospector vote – those who ‘vote pragmatically for whichever party they think will improve their financial circumstances’ – has much resonance. The grey middle class might not be the sexiest of demographics, but they often decide elections. Miliband may have gained 12 per cent more 18-24 year olds (turnout 43 per cent) in 2015 than five years earlier, but the fact that he managed to do 8 per cent worse than Gordon Brown’s 2010 performance with the crucial over 65s (turnout 78 per cent) put the final chisel in the Edstone.

Perhaps if you give young voters a “radical alternative” they really will turn out – though worth recording that turnout amongst under 25s at the ‘real choice’ election of 1979 was the lowest either side of the majority Labour governments of 1966 and 1997 – but there are no guarantees. All this is a challenge for Labour per se however, not just Corbyn.

For the bookies’ favourite himself there are some specific complications. Big ticket policies like People’s Quantitative Easing have been queried by fellow leadership candidates (to declare an interest, while I am a Kendallite, I wrote a report arguing for a much truncated, one-off form of People’s QE in 2012), though it is just about backed by councillors in our survey. Corbyn’s foreign policy choices of threatening to leave NATO (rejected by two thirds of councillors) and scrapping Trident (rejected by a third) are also likely to be controversial. And the sum total of a left leaning agenda – as Ed Miliband discovered – is often less than its constituent parts. If Jeremy Corbyn is going to become the first opposition leader since 1906 to gain a full parliamentary majority whilst pledging to raise the top rate of income tax, he’s got a lot of work to do.

But our survey suggests that he’ll get the time to do it. If our data suggests Corbyn is at present unlikely to be Prime Minister, for all the talk of an early coup against him, he looks in a strong position to at least contest that election. And that remains an astonishing rise.

Richard Carr is a Lecturer in History at the Labour History Research Unit (LHRU), Anglia Ruskin University. The LHRU has today released new polling data on the Labour leadership. The views expressed in this article are not necessarily those of the LHRU, the kind councillors of all parties who took time to answer the survey, or Anglia Ruskin University.

Our Partners