Home

LOTSOFMONEYBALL

Published on September 12th 2017, by Johnny Phillips

The Moneyball Story

Twenty years ago Billy Beane took over as General Manager of the Oakland Athletics baseball team. The once mighty franchise, under new ownership, were now constrained by one of the smaller budgets in Major League Baseball. Frustrated at always being beaten to the signings of the best players by richer rivals, Beane turned to using data (known as Sabermetrics in baseball) to find under-valued players he could sign. He figured that if he could identify players who were actually better than their going market rate – and that the A’s could therefore afford – then he could hope to compete with, and even beat teams who paid more for their talent. The idea of Moneyball was born.

Michael Lewis wrote a best-selling book based on Beane and his A’s, and in particular their 2002 regular season – in which they played well. The book was then turned into an Oscar nominated film starring Brad Pitt, and legend of Beane was sealed. And Moneyball became a catch-all term for stats geeks using data and computers to get an edge on old-school ‘gut-trusters’ in all sorts of walks of life, and particularly in professional sports. Moneyball promised that the geeks would inherit the earth.

The Moneyball Myth

But they didn’t. Twenty years on and Billy Beane is no closer to winning a World Series championship with the A’s than he was when he took over. It is a myth that he found a stats-based way to ‘win an unfair game’. In fact he’s only got past the first round of the postseason on one single occasion, and that was over a decade ago And in the last three seasons the A’s have been one of the worst teams in baseball, finishing bottom of the AL West in ’15 and ’16 (and are on course to do the same again this year). Moneyball was a brilliantly written, and engaging story. But it’s a myth. Billy Beane did not turn the Oakland A’s into a champion baseball team using Sabermetrics (aka analytics). He didn’t master the art of winning an unfair game. There is no more evidence that a ‘Moneyball approach’ can help a professional sports team consistently beat the hierarchy imposed by money than there is that the Loch Ness Monster exists.

It’s true that the A’s did have a brief upturn in fortunes back in the early 2000’s (including the twenty game winning streak that was central to Lewis inventing the Moneyball story) but that proved short-lived. If the use of analytics conferred any competitive advantage then it didn’t last long. Which is logical. Because of course there was never any barrier to rich teams using analytics too. This is very different to the world I usually inhabit – professional sports betting  – where I can find a consistent edge using analytics to analyse sports and build betting models for the high-staking professional gambler clients who use the output. This is partly because the vast majority of sports bettors are  ‘gut-trusters’ who don’t use analytics (or who use it badly).

The Alternative Story

The Oakland A’s DID have three excellent regular seasons back to back from 2001-3. But a rather less romantic potential explanation for this emerged in recent years, when two of the star players from their Moneyball team (Jason Giambi and Miguel Tejada) were found to have been taking illegal performance enhancing steroids during that time, while their team-mate Adam Piatt confessed to being a supplier of steroids. Maybe the A’s found an edge by doing drugs, not doing stats?

Coincidentally, in 2002, at the same time the A’s were enjoying back-to-back 100+ game regular season wins, Lance Armstrong was in the middle of his streak of winning 7 Tours De France. And when Moneyball made its way to the top of the decade’s best-sellers list it joined Armstrong’s Its Not About The Bike. Nobody is falling over themselves to give Armstrong a load of sympathy these days of course, but it’s probably a bit harsh that his book is now listed by Wikipedia as ‘Fiction’ while Moneyball is accepted as ‘Fact’.

It’s a depressing but unavoidable reality that in an age where modern medical science is capable of enhancing human athletic performance beyond the limits of hard work, lots of sleep and a good diet, many sporting fairy-tales end up being revealed to have a darker truth lurking behind the glossy veneer.

The Reality

In a mature and efficient labour market such as exists in sports like cycling, baseball and soccer Moneyball doesn’t work. There is no such thing as the ‘art of winning an unfair game’. Teams are (give or take a little, and allowing for the odd short-term outlier) as good as the sum of the talent of the players they contain. And the market in athletic talent in these sports is efficient. So the best players command the biggest salaries, which only the richest teams can afford. So they become the best teams. This is as true with the New York Yankees in baseball, Team Sky in cycling as it is with Real Madrid, Barcelona, Bayern Munich, Juventus, PSG, Celtic and England’s ‘Big 5’ in their domestic football leagues. The size of the superiority that the richest teams in football leagues have over their poorer correlates (virtually) perfectly with the difference in the size of their salary budgets.

Performance enhancing drugs are perhaps the only consistent way to disrupt this reality, because unlike ‘marginal’ gains, they have the potential to alter the fundamental natural athletic capacity of a sportsman. So cheap doped players can perform as well as non-doped expensive ones (until they become expensive too).

Marginal gains are insignificant compared to the fundamental differences that exist in athletic capability, which can be bought in an open market. What really matters to long-term success of sports franchises like football clubs is how much money they have to spend on talent. Plus, crucially: Strategy (the art/science of long-term planning) not least because this is how a club can maximise the amount of talent that it can acquire with its salary budget.

The Theory

So says me. The whole world thinks managers, formations and tactics matter in football. I say they don’t. Football teams are as good as the sum of talent of the players in them. That’s my theory. I say that the salary that a football player commands is a good approximation of his talent. So we can measure how good a football team will be based on how much it spends on player salaries. We don’t need to know who the manager of a team is, what formation they will play or what tactics they will adopt. We don’t even need to know who the players are. We just need a decent guide to how good they are. And we can model that if we know how much they get paid. Let’s call it the lotsofmoneyball theory.

The Experiment

As the labour market in European professional footballers is open and subject to market forces, if we are right about the lotsofmoneyball theory then we should be able to model and predict the Champions League group stage. It brings together 32 teams from 17 different countries. And thanks to Nick Harris and the excellent www.sportingintelligence.com website I can get income figures for each club to put into my model.

The Competition

The group stage of the Champions Leagues features 8 groups of 4 teams, who each play one another home and away. Six games isn’t a huge sample size, but it’s enough to generate a ‘Points For’ and ‘Goal Difference’ value for each team that we can use. The knockout stage of the competition is subject to more randomness with random draws for each round, and the vagaries of deciding a winner of each tie over just 180 minutes. But even then, you would struggle to do better than guess that the richer team would win each knockout round tie too.

The Model

As a proxy for the size of each team’s salary budget we have used the latest annual income figures of each of the 32 Champions League group participants, as reported by Sporting Intelligence. Our model takes these values and converts them into a prediction for points and goal difference in each of the 4-team groups. And that’s it. The model doesn’t know or care who the team is, which country they come from, the identity of their manager, where they finished in the league last season, their current form/injuries/suspensions. All it needs to know is how rich the club is. It’s a single input model. With one exception.

The England Tax

I am proposing the lotsofmoneyball theory as the rule that governs Champions League football. As with every rule, there is an exception though. The Income Model method of predicting performance in the UCL works for teams from every country playing in Europe, apart from one. The exception is England. In recent years England’s Premier League top teams have got worse. At the same time as they have been getting richer faster thanks to Sky and BT TV money. This is weird, but the evidence is undeniable. Why should it be?

Screen Shot 2017-09-12 at 11.33.38

Screen Shot 2017-09-12 at 11.30.34

I have a few theories; like the idea that top English teams have made the greatest Moneyball investment in the use of analytics than elsewhere, but (with the honourable exception of Southampton) they have been really bad at it, so have actually made themselves worse. British clubs are virtually alone in persisting with the inefficient management structure that sees an all-powerful ‘manager’ given total responsibility for the football operations at a club. This can lead to ruinous short-term thinking, and damaging upheaval when that managers changes (as they do regularly). This makes for terrible strategic planning, typified by the likes of the jaw-droppingly idiotic situation where an English Premier League team can hire and then sack a new manager just four games into a new season.

Crystal Palace Managers, last 10 years.

Crystal Palace Managers in the last ten years. The very definition of woeful strategic planning.

Points per game in English league football

And the end result of woeful strategy

There is also the possibility that, as with many hard-to-explain peaks or troughs of sporting performance, history will eventually reveal that the use of performance enhancing drugs plays a part. It’s possible (although arguably potentially libellous – so I’m not going to say ‘probable’) that English clubs have fallen behind some of their continental European rivals because they haven’t been exploiting the latest PED aids to athletic performance that are possibly/allegedly/probably being widely used in other countries.

Whatever the reason, the lotsofmoneyball model massively over-rates English teams. So we impose a ‘40% English Tax’ on them. In other words, we credit them with only 60p in the £ ‘value’ for their total income in our model.

The Predictions

So here is the output of our lotsofmoneyball income based Champions League group predictions. There are a few anomalies – like Monaco, who benefit from royal patronage and an in-built tax advantage to offset low home crowds – and questionable income figures (the term ‘financial doping’ is becoming mainstream, as oil-rich nation states replace oligarchs as the biggest boys in the European football playground) but other than the England Tax these are pure output values from the model.

Income Model UCL 17_18

A prediction of Points and Goal Difference for the 8 Champions League groups in 2017_18 made before the first games.

The Challenge

Your challenge is to see if you can beat me. Should be easy, right? I have only used a single input for my very simple model. You are free to use whatever you like – your knowledge of teams, players, managers, tactics, formations, fixture schedule. Latest results. Computerised rankings. Predictive models –anything you like. But I like my theory and my prediction, that took me 5 minutes to generate.

Let’s see if you can beat it. Fill in your predictions for the 32 teams in the Champions League group stage and post it on Twitter using #lotsofmoneyball . Tell us how you came up with your prediction. Once the Champions League group stage is over we can then measure how good everyone’s predictions were. You score ‘1 point’ for the difference between your prediction and the actual result for Points Won and Goal Difference, for each of the 32 teams. Lowest score wins (i.e. a prefect set of predictions would score the ‘best’ score of 0).

Good luck!

 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

w

Connecting to %s