Sports Book Chat: Chris Anderson

In the beginning, there was Moneyball.

Well, not really.  What kicked off sport’s interest in statistics was an underground movement in baseball headed by Bill James and which eventually made its way into the mainstream in the form of statisticians being employed by clubs. It was James who coined the term ‘sabermetrics’ (the search for objective knowledge about baseball) and who started publishing his theories in the highly influential Baseball Abstract books.

Yet it was Michael Lewis’ book ‘Moneyball’ about the Oakland ‘A’s’ use of statistics to make up for their financial disadvantage that brought it all to the public consciousness. Since then anyone wanting to try and use statistics in any sport has had to live with claims of taking a ‘Moneyball approach’.

And no sport has had to hear such accusations of late as much as football. All of which makes it all the more surprising that there is still a huge misconception as to what ‘Moneyball’ actually is. Indeed, many still harbour the belief that it is all about using statistics to determine which player should be signed and which shouldn’t. Or to use numbers to determine which players to use or which tactics to adopt.

But it isn’t about that. Or, at least, not only about that. It all centres around the idea of using statistics to gain an edge or to confirm whether the cold facts that are held in the numbers can support widely held perceptions.

Take corners. There is an inevitable rise in anticipation every time a team wins a corner, which is understandable given that this leads to a greater chance of scoring. Except it doesn’t.

That is what Chris Anderson and David Sally argue in their book ‘The Numbers Game – Why Everything You Know About Football Is Wrong’. Again, this being a book about sports and which uses numbers, it is somewhat inevitable that ‘Moneyball’ gets a mention, but this is much more than that: whilst Moneyball was trying to explain a movement, The Number’s Game simply want to help you understand football better. Chris Anderson kindly agreed to this interview:

First of all, what’s your background? Am I right in thinking that you’ve got a past as a goalkeeper?
Yes, you’re right. I was born and raised in the old West Germany. The 1974 World Cup came around and inspired a lot of kids to take up football. My friends and I enacted the matches in an alley near my house. I ended up keeping goal for a number of years after that; eventually for a local 4th division side. When I realised I was not cut out to make it as a professional, I hung up my gloves and went to university instead, eventually earning a PhD and becoming a professor at an Ivy League university in the United States. My academic career has mostly focused on an area called political economy.

What do you make of the growing interest in the use of statistics in football? Are you surprised about how rapidly it has progressed?
We’re not really surprised. It’s part of a broader trend in professional sports. In the wake of Moneyball and the story of the Oakland A’s, analytics changed the face of baseball in the U.S. – how players are recruited, coached, and so on. In basketball, the NBA, too, is currently in a frenzy of statistical advances as clubs leapfrog each other to hire young talented quants and employ the latest technology. Lagging a little behind are the NHL (ice hockey) and NFL (American football) clubs. It’s not just an American phenomenon – think about the efforts made by Team GB for instance; rugby clubs are interested in analytics, and so are cricket and tennis. There is every reason to think that all sports in which the competition is fierce and the financial resources are adequate will go down the same path.

Then there is a push from the supply of data – compared to the days when Charles Reep, the man we call the match accountant and one of the heroes of our book, hand-coded events on the pitch, the costs of acquiring match data have plunged by orders of magnitude. When data become cheaper, more people will try to use them.

Finally, the advances in computing power and storage and statistical software that have affected all organisations in this new era of Big Data should have the same impact on football clubs. In that sense, football is no different from Tesco.

So in the end, we think it’s a kind of natural progression that will only speed up, not slow down, though the path it’ll take will be uneven and hard to predict.

What made you decide: let’s write a book?
It seemed like a fun change from writing another book on political economy! But more seriously, given the trends mentioned before, we thought someone should write a book that captures this moment of change in football, and perhaps provides a bit of additional momentum behind the reformation of numbers and statistics in the game. We call it a reformation because the revolution already happened – people like Reep and Charles Hughes and Stan Cullis and Lobanovskyi and Sam Allardyce and others were the vanguard of revolutionaries, though some of them failed miserably. In many ways, writing the book was also a journey of discovery – we were intensely curious about what was being done in football and what the numbers can and cannot tell us that we don’t know yet. We were fascinated by what we found. Turns out, there’s a lot going on in the game and behind the scenes; some of it is quite technical, some of it very political. For us, the challenge was to write about it in a way that’s fun and engaging – a book we ourselves would want to read – and we hope we did that.

To those who have read ‘Soccernomics’, why should they read ‘The Numbers’ Game’? What makes it unique?
We loved Soccernomics; along with Jonathan Wilson’s Inverting the Pyramid, it was one of our inspirations in writing The Numbers Game. We thought the application of economics and political science to football with its examinations of topics like national wealth and national team performances, supporter suicides, and discrimination in the labour (transfer) market, for example, were really informative and creative. Aside from the penalty kick chapter, though, we felt that much of the action on the pitch and on the training ground was still left to be examined, and that is what we do in The Numbers Game. To over-simplify, our book is more focused on the match, and Soccernomics is more focused on the sport. So we think of these books as being complementary; we’re thrilled you’d mention them in the same breath.

How did you decide what you were going to tackle?
That was harder than we thought. We started with the question about Rory Delap’s throw-ins and Stoke, and we just kept asking questions about what goes on in a match and in football clubs, and why. We then asked what data or research we could get our hands on that might be able to provide some insight to produce an answer. Inevitably, a tentative answer would end up raising even more questions. At some point, we had a list of about 25 chapters. We then whittled it down, but even that was tricky. So our decision rule was to focus on some of the most fundamental questions rather than of the more technical issues. For instance, is football a game of skill or luck? What matters more: offence or defence? What is possession? How would you build a team? How could we know if managers matter? And so on – the stuff we thought everyone should or would want to know. Unfortunately, this meant we had to leave some really fascinating stuff to the side. But to tell the truth, our original manuscript was more than twice as long as the final version – there was so much to write about. Our editors, however, felt that an 700-800 page tome on football numbers might not be the best idea from a marketing perspective, so we had to ruthlessly cut down each chapter to the length that they are today.

What was the most surprising area you ended up writing about?
We were convinced, as many people are, that football morphs and mutates as it crosses borders. German football is different from Spanish is different from English, etc. The numbers don’t really support this view. For the measures that matter the most – the things that decide games – the football played at the very top, in the very best leagues in the world – are basically indistinguishable. Essentially, when it comes to goals, shots, penalties, corners, you name it – it’s the same game in Premier League as it is in Serie A, La Liga, and the Bundesliga – even though it may appear to be different. The question, of course, then is why? As we say in The Numbers Game, we think this convergence is the natural effect of competition on the diffusion of talent and football knowledge across the globe.

You’ve been to a lot of clubs looking at how they work.  What did you find? How prevalent is the use of statistics? How much did the football people at the clubs believe in those statistics? Or was it something that they did because everyone else was doing it?
It’s really a mix of everything. Some clubs have made significant investments in the infrastructure required to make evidence-based decisions in various areas of the club – a good example is Manchester City – while others have barely begun or have decided not to go down that road at all. Our sense is, generally speaking, that clubs typically want the numbers to tell them some very specific kinds of things, while others don’t really know what they want. Right now, a big interest is in the area of recruitment – tell me which player is better or how much a player is worth. Then there’s also statistical work that’s done during the process of due diligence when signing a player. Another area in which data are playing a growing role is in sports science – the physical and physiological side of the game. Here, there’s more of a focus on monitoring and benchmarking performance.

At the same time, there are plenty of people working for clubs who don’t quite know what to make of all the data emerging in football but want to get in on it. They’ve read Moneyball or heard of the MIT Sloan analytics conference, but the clubs they work for don’t really have an infrastructure that helps to collect, digest, and communicate quantitative information, and they don’t always have good ways of building it into their day-to-day schedules. For them, it’s a steep learning curve in terms of how to think about data, stats, software, what have you, and what it’s useful for and how to incorporate it into the club’s structure. Often, they lack a champion or support from the top.

There’s a lot of great work being done by bloggers such as Paul Tomkins and Dan Kennett but is this really ground-breaking stuff or do the statisticians employed by the clubs do similar work but we’re simply not aware of it?
Every club subscribes to at least one data provider for match data, for instance – be it Opta, Amisco/ProZone, StatDNA, etc., and every club monitors players’ training and conditioning, and so on. So clubs are quite similar in the availability and supply of the numbers they can work with, and the numbers have become plentiful and easy to access. They do differ in terms of the level of resources, manpower, and attention they pay to the data, however. On the match and scouting data side, it is fair to say that the typical club uses numbers as basically “enhanced video analysis” – that is, to put some numbers to trends that scouts would have previously had strong impressions of.  There is less attention paid to whether these now quantified trends matter – do they affect the chances of winning the match? Do they mean one player is better than another – and there are challenges in translating the analytical knowledge into reality through the manager and the players.

By and large, football clubs form an “island world” in which there is a lot of imitation and mimicking; some note is clearly paid to what’s going on outside, whether it be to insightful bloggers or to other sports, but it’s all driven by the practicalities of doing a job and realities inside a club. We’ve certainly come across a number of people inside clubs who read the blogs and talk about what they see being done by outsiders. Sometimes, work by outsiders is useful, sometimes it isn’t. The trouble is that it’s quite general when the questions being asked in clubs tend to be more specific.

When Rafa Benitez was talking about the need to build a strong squad at Liverpool, he didn’t get a lot of compassion especially as there was the feeling that he had a strong enough starting 11.  Yet you’ve proven that the strength of your fringe players is a significant marker of success, right?
That’s correct. As we say in The Numbers Game, we believe that football is in fundamental respects a weakest link game – the simple version of that is that our weakest player is as important, if not more important, than our strongest player. Although we focus on individual players as specific weak links, it’s important not to take that too literally; the weak link could also be a serious lack of understanding and mis-coordination between two or more players on the pitch. Going back to the Benitez example, it’s critical to have a deep enough and varied enough squad because your strongest 11 aren’t always available or who your strongest 11 are actually depends on your opponent or the time of year or any number of other factors. I was recently reminded of that when reading the really excellent Champions League Dreams by Rafa and Rory Smith.

It’s also important to remember that this does not mean that your strongest players, a Suarez, a Gerrard, don’t help you win, because they certainly do. But it is to say that the relative quality of your weaknesses has an even bigger impact on how many points you secure in a match and in a season than does the relative quality of your superstars. United last season might be a good example: they maintained their success rate in the latter half of the season even as van Persie’s production seemed to decline. Why? Because of the quality that ran down through the tail end of the starting eleven and the usual substitutes. Or think of Arsenal with and without van Persie – in the last season with him, they won 70 points; without him they produced 73.

Let’s lay out a realistic situation: if Liverpool are forced into selling Luis Suarez, in what way could statistics be used to help them find substitutes who end up doing more for the team?
The key is in your question. Individual players come and go, but football is a team game. So the trick is to figure out what a team needs to do to win games in the Premier League, and then to see how the individual and sets of players you have or might get will help you produce the right team performance up and down the pitch. Statistics can help the club identify the key markers of team production they are trying to maximise; and different players are likely to have different strengths to get the team there. Data can help you diagnose your team’s needs and possible ways of filling them with individual players.

England is perhaps unique in that a corner is met with almost as much enthusiasm as a goal (well, everywhere except at Liverpool because we seem incapable of taking them).  Yet you argue that that enthusiasm is misplaced, don’t you?
We do: turns out the number of corners your team gains has only a tiny correlation with its success in a match. Although the ball is quite close to the goalmouth geographically as it’s placed in the corner, statistically it is much further away than it seems. Ironically, one reason for that of course is that the defending team thinks corners are dangerous, so they defend them systematically and vigorously, further diminishing their impact. What is also under-appreciated is that your own chance of scoring is balanced by your vulnerability to a counter-attack. Just two days ago, we had an email from a performance analyst for a top club in Brazil who was reading The Numbers Game and had found the same in Série A; it seems to be a general phenomenon.

What’s next for you?
We are continuing our work on football analytics, and we are hoping to gain even more insight into the inner workings of this wonderful, complicated, challenging sport.

This interview originally appeared on The Tomkins Times.