Authors: Simon Kuper,Stefan Szymanski

Tags: #Psychology, #Football, #Sports & Recreation, #General, #Self-Help, #Social Psychology, #Personal Growth, #Soccer

Soccernomics (6 page)

SHOULD DO BETTER:

IS ENGLAND WORSE THAN IT OUGHT TO BE?

That England underachieves is usually taken for granted in the British media. After all, the team hasn’t won anything since 1966, and sometimes doesn’t even qualify for tournaments. Clearly, that is not good enough for

“the mother country of soccer, birthplace of the greatest game.”

But does England really underachieve? Or is it just that the English expect too much of their team? To answer this, we first need to work out how well England should do,
given its resources
.

Before we are accused of looking for excuses, let’s consider what is and isn’t possible. A five year old can’t win the hundred-meter finals at the Olympics, and neither can a seventy year old. You aren’t going to have a career in the NBA if you are only five feet tall, and you’ll never ride the winner in the Kentucky Derby if you are six foot eight. It is very unlikely that you will have a career in show jumping if your parents earn less than forty thousand dollars per year, you probably won’t win a boxing match if you’ve never had any training as a boxer, and you won’t have a shot at being world chess champion unless you can persuade a team of grand masters to act as your seconds. Genetics are beyond our control; training depends partly on our own effort, but partly on the resources that other people give us.

What is true for the individual is also true for the nation. During his tenure an England soccer manager cannot easily (a) increase the size of W H Y E N G L A N D L O S E S A N D O T H E R S W I N

the population from which he will have to draw the talent, (b) increase the national income so as to ensure a significant increase in the financial resources devoted to developing soccer, or (c) increase the accumulated experience of the national team by very much. (England has played more than eight hundred games since its first game in 1872, and currently plays around a dozen games per year, so each extra game doesn’t add much to the history.)

Yet in any international match, these three factors—the size of the nation’s population, the size of the national income, and the country’s experience in international soccer—hugely affect the outcome. It’s unfair to expect Belarus, say, to perform as well as much larger, more experienced, and richer Germany. It is fairer to assess how well each country
should
perform given its experience, income, and population and then measure that expected performance against reality. Countries like Belarus or Luxembourg will never win a World Cup. The only measure of performance that makes any sense for them is one based on how effectively they use their limited resources. The same exercise makes sense for England, too, if only as a check on tabloid hysteria: does England really underperform given what it has to work with?

In absolute terms England is about tenth in the world. But we want to know how well it does in relative terms—not relative to the expectations of the media but relative to English resources. Might it be that England in fact
overachieves
, given the country’s experience, population, and income?

To work this out, we need to know the soccer results for all the national teams in the world. Luckily, we have them. There are a number of databases around of international matches, but one of the best is maintained by Russell Gerrard, a mathematics professor at Cass Business School in London. By day Russell worries about mathematical ways to represent the management problems of pension funds. For example, his most recent paper is snappily titled “Mean-Variance Opti-mization Problems for the Accumulation Phase in a Defined Benefit Plan.” It concerns, among other things, Lévy diffusion financial markets, the Hamilton-Jacobi-Bellman equation, and the Feynman-Kac 34

representation. As you might expect, Russell has been meticulous in accumulating the soccer data, which took him seven years. His database runs from 1872 through 2001, and includes 22,130 games.

RUSSELL GERRARD’S BRILLIANT SOCCER DATABASE

Later in the book, we will crunch Russell’s data to discover which is the best soccer country on earth, and which punches most above its weight.

But here, let’s limit ourselves to a sneak preview of where England stands.

The distant past is of limited relevance. Let’s therefore concentrate on the most recent period of Russell’s database, 1980–2001. This was by no means a golden age for England. The country didn’t even qualify for the European championship of 1984, or for the World Cup ten years later, and its best moments in these twenty-two years were two lost semifinals. Altogether in the period England played 228 matches. It won 49 percent of them and tied 32 percent, for a “win percentage” of 65 percent (remember that for these purposes we treat a tie as half a win). That is near the middle of the country’s historical range.

We want to see how much of a team’s success, match by match, can be explained by population, wealth, and experience. However, win percentage is not the best measure of success, because any two wins are not the same. We all know that a 1–5 away win against a certain someone is not the same as a tame 1–0. Put another way, if England plays Luxembourg and wins only 1–0, it’s more likely that the Luxembourgian press will be in ecstasy than the British tabloids.

Instead, we chose goal difference as our measure, since for any match we expect that the greater the difference between the two teams’

populations, wealth, and experience, the greater will be the disparity in scores. (Of course, a positive goal difference tends to be highly correlated with winning.)

We then analyzed Russell’s database of matches using the technique of multiple regression. Quite simply, multiple regression is a mathematical formula (first identified by the mathematician Carl Friedrich Gauss W H Y E N G L A N D L O S E S A N D O T H E R S W I N

in 1801) for finding the closest statistical fit between one thing (in this case success of the national team) and any other collection of things (here experience, population, and income per head). The idea is beautifully simple. The problem used to be the endless amount of computa-tion required to find the closest fit. Luckily, modern computers have reduced this process to the press of a button (just look up “regression”

on your spreadsheet package). For each international match you just input the population, income per head, and team experience of the two nations at that date, and in seconds you get a readout telling you how sensitive (on average) the team’s performance is to each factor. We will also take into account home advantage for each match.

Collecting the data is usually the toughest part. We have assembled figures for the population, soccer experience, and income per capita of 189 countries. We will unveil our findings about the other 188 countries later in the book. Here, we will focus just on England and its supposed underperformance.

We ran our regression and immediately made several discoveries about international soccer. First, home-field advantage alone is worth a lead of about two-thirds of a goal. Obviously, that is nonsense if applied to a single game, but think of it this way: playing at home is like having a goal’s head start in two out of every three games. Second, having twice as much international experience as your rival is worth just over half a goal. In fact, experience turns out to matter much more than the size of your population, which is why the Swedes, Dutch, and Czechs do better at World Cups than the very large but inexperienced US. Having twice your opponent’s population is worth only about one-tenth of a goal. Having twice the gross domestic product (GDP) per head is worth about as little. In other words, although being large and rich helps a country win soccer matches, being experienced helps a lot more. That is not good news for the US.

It should be added that our estimates are statistically very reliable.

Not only is there not much doubt that these factors matter, but there is little doubt about the size of the effects. It is these effects that make the first rounds of World Cups fairly predictable.

However, much still remains unexplained. Experience, population, and income per capita combined explain only just over a quarter of the variation in goal difference. That is good news: if we could predict outcomes perfectly by just these three factors, there would not be much point in watching World Cups at all. Nonetheless, the fact that these three factors explain so much tells us that, up to a point, soccer is rational and predictable.

For now, we are interested only in England. Are its resources so out-standing that it should do better than merely ranking around tenth in the world?

First let’s look at experience. England is one of the most experienced countries in soccer. It played 790 internationals between 1872

and 2001. According to Russell’s data, only Sweden played more (802).

However, England’s much vaunted history is not worth much against the other leading soccer nations, because most of them have now accumulated similar amounts of experience. Brazil, Argentina, and Germany had all played more than 700 internationals by 2001. When it comes to our second variable, national income, England scores high, too. It is usually one of the richest of the serious soccer countries.

Where England falls short is in size. What often seems to go unnoticed is that England’s population of 51 million puts it at a major disadvantage to the countries it likes to measure itself against in soccer. Not only is Germany much bigger, with 80 million inhabitants, but France and Italy have around 60 million each. Among the leading European nations, England is ahead only of that other “notorious underachiever,”

Spain (40 million). So in soccer terms, England is an experienced, rich, but medium-sized competitor.

Then we ran the numbers. We calculated that England, given its population, income, and experience, “should” score on average 0.63

goals per game more than its opponents. To get a feel for how this works, consider England’s performance at the World Cup of 1998.

England played Tunisia, Romania, Colombia, and Argentina, all on neutral ground, so there was no “home-field” effect. Facing Tunisia in its opening game, England had a population five times larger, a GDP

W H Y E N G L A N D L O S E S A N D O T H E R S W I N

per head four times larger, and two times more international experience than its opponent. That combination gave an expected goal difference of one goal in England’s favor. In the event, England won 2–0 and so did better than expected.

Against Romania, England had the advantage of twice the population, five times the income per head, and a bit more experience (other countries have been playing soccer for longer than the English sometimes like to think). All this gave England an expected advantage of about half a goal. England’s 1–0 defeat meant that it underperformed by one and a half goals.

Next England played Colombia, with a slight advantage in population, four times the income per head, and double the experience. The package was worth an advantage of almost an entire goal (you start to see how the Tom Thumb World Cup might be organized), but again England overperformed, winning 2–0.

Then came Argentina: as everyone in England knows, the English lost on penalties. Should England have done better? Well, it had a slight advantage in population over Argentina, double the income per head, but slightly less experience (Argentina plays a lot of games). Putting all that together, a fair score would have been a tie, which is exactly what happened after 120 minutes. Pity about the penalties, but more about those later in the book.

Our model allows us to reexamine every game ever played. The glorious uncertainty of soccer means that there are many deviations from expectations. However, if we average the difference between expectations and results for each country, we get a picture of whether any national team systematically outperforms or underperforms relative to its resources. Our finding: England in the 1980–2001 period outscored its opponents by 0.84 goals per game. That was 0.21 more than we had predicted based on the country’s resources. In short, England was not underperforming at all. Contrary to popular opinion, it was overperforming.

As an example, take England’s games against Poland. England played the Poles eleven times in the period, winning seven and tying four times, with a goal difference of plus sixteen. Over the period England’s 38

population was about 25 percent larger than Poland’s, its income per head about three times greater, and its international experience about 20 percent more. These should have contributed for a positive goal difference of about one, three, and one goals, respectively, or a total of plus five. So England’s goal difference of plus sixteen was eleven goals better than we might have expected. That is not too shabby.

Later in the book we will reveal our global table for relative performance from 1980 through 2001: a ranking of the teams that did best relative to their countries’ experience, income, and populations. For now, we’ll just say that England came in 67th out of 189 countries. That put them in a group of moderate overachievers, just below Russia, Azerbaijan, and Morocco and just above Ivory Coast and Mozambique. Like England, all these teams scored about a fifth of a goal per game more than they

“should” have, given their populations, income, and experience.

However, England doesn’t benchmark itself against Azerbaijan. It’s more interesting to see whether England overperforms “more” than the teams it sees as its rivals: the best countries in the world. Here’s a ranking of the game’s giants, plus England, based on how many goals per game each scores above expectations:

“Additional” Goals per Game

Country

Above Expectations

Brazil

0.67

France

0.35

West Germany

0.28

England

0.21

Italy