AFL Analytics: Part One - Moneyball Is Old News
Is it time to move on from Moneyball?
Let me explain.
When Michael Lewis wrote about Billy Beane and the Oakland Athletics in 2003, the notion of sports analytics was relatively new. Harnessing market inefficiencies was the cornerstone of the business world - of Wall Street and boardrooms, not stadiums and locker rooms.
Sabermetrics, pioneered by stats guru Bill James and now understood as the much more digestible “Moneyball” thanks to Lewis’ book, is defined as the empirical analysis of baseball data, originally used to identify players in the marketplace who seemed to be undervalued.
For a poorly-resourced team like the A’s, this was a chance to look at baseball through a different lens and gave them a way to compete.
It is fast approaching twenty years since that book was released. Nearly ten years since Brad Pitt and Jonah Hill gave the story mainstream exposure on the big screen, with $110M of gross profit to go along with it.
Yet in some ways we haven’t moved past Moneyball as just a buzz word – a way to talk about sports data, even without a great grasp of how it is used by the teams.
For the baseball world, Moneyball was just the beginning of what is now the new normal. Analysts don’t look at all out of place in the office (they also probably don’t look anything like Jonah Hill, either), and using numbers to determine recruiting strategies, trade targets and lineups is just part of good sports business practice.
Sport advances incredibly quickly, especially in pursuit of an edge. On the brilliant Trailblazers Podcast, Billy Beane lamented the release of Lewis’ book had quickly exposed their competitive advantage, so less than two years later they “had to go back to the drawing board”. Almost as soon as they had discovered their edge, it was gone.
Naturally, the concept infiltrated other sports soon after.
The NBA embraced the analytics revolution and the sport has forever changed as a result. In the NFL, some teams have unwittingly provided a great example of how an analytical approach can backfire spectacularly when the implementation fails. Closer to home, even the Big Bash has invested heavily in data thanks to Cricket Australia’s deal with CricViz.
But even in baseball there have been varying degrees of success, and a grey area around how to strike the right balance between the old and the new.
The Boston Red Sox won the World Series using Moneyball principles in 2004, 2007 and 2013. But after three poor seasons, they made a conscious decision to move away from the data. Owner John Henry said at the time, “we have perhaps overly relied on numbers.”
Elite sport has a great habit of making you question previously sound decision-making when you’re losing. And perhaps not be as critical when the wins are on the board.
So what about Australian Football?
As far back as 2006, RMIT conducted detailed research into how Moneyball principles could be applied to the AFL. So long ago, in fact, that the company who supplied the statistics for the study is no longer in business, and hasn’t been for years.
The most important point came in the paper’s conclusion:
“Australian Football is a relatively complex sport given its free-flowing structure. This is quite distinct from baseball, where a game is essentially composed of a series of discrete transactions.”
Baseball: pitcher vs batter. Cricket: bowler vs batsman. Basketball changed dramatically on the back of shot location data. The closest thing in AFL might be set shot goalkicking – player vs the goalposts, perhaps. Our sport is incredibly unique.
There are so many reasons why any future data revolution will look far different in the football world. Among them:
- The number of players on the field
- Play is continuous, with no time outs
- Restrictions on runners and in-game messaging
- The need for player consent when trading
- A salary cap AND a spending cap
- Very few pure 1v1 contests outside of ruckmen at stoppages
If the analytics really took over, could we see some players refuse to take set shots from certain locations, preferring to pass to a teammate even though they may seem to be in a worse position to the eye?
Highly unlikely, but could set shot selection eventually come down to a pure “percentage play” based entirely on historical conversion numbers? Could teams go without a wingman if there’s only a 2% chance of the ball heading there after a centre bounce? Or would they do away with entire centre bounce combinations because the numbers are telling them their collective clearance rate is through the floor? Time will tell.
The one fundamental similarity between all sports is the challenge of embracing evidence-based practice, and how far teams want to take it. “Adapt or die” was the simple line delivered by Brad Pitt in Moneyball. Similar real-life conversations wouldn’t be uncommon. If everyone isn’t on the same page, the balancing act between traditional methods and analytics can fail before it’s even started. Like the Cleveland Browns, for example.
For the data “nerds” (there will be a time when they are shown a little more respect to shake that tag, but perhaps not just yet), humility is the key.
Champion Data stalwart turned North Melbourne List Manager, Glenn Luff, summed it up perfectly:
"It's funny – people think, 'Oh, Moneyball, Luffy's just going to throw data at everyone’.”
"I'm not silly enough to think it's just data. Footy is the hardest game in the world to capture statistically but it's a piece of the puzzle that I think you need to understand.”
Just as a coaching group or recruiting team don’t have all the answers, a data scientist or analyst can’t pretend to own the secret to success. With that attitude inside a team, you’ve probably already failed.
Everything is Moneyball!
In contrast, the media have been a little more easily seduced, with a tendency to apply the Moneyball lens to everything. Granted, it does make for a catchy headline.
Often the easy route is to use the term as a lazy way of describing any use of data, even though it can often bear little resemblance to the original ideas of Bill James, Billy Beane and co:
Target some guns in the free agency and trade period - Moneyball.
Retrospectively look back on anyone who was picked up by another club in the past 5 years – Moneyball.
A football-loving Minister Of Parliament with an interest in economics? – Moneyball.
At least this article uses Bill James mathematics to try and predict the 2020 ladder – Actual Moneyball.
Of course there are also some great insights into how far along the journey we are, such as this piece from The Guardian. The publication of Footballistics and the restricted access of some more advanced football metrics is something we will tackle in Part Two.
Ultimately, it’s the secretive nature of clubs (and rightly so) that contributes to an under-educated AFL media and general public about what measurements are useful internally. But we also face the problem that the numbers used by teams don’t often make for good reading.
Sharp minds at club-land have been employing Moneyball principles and advanced use of numbers for many years, particularly in the recruiting space – they just wouldn’t be telling anyone what they’re doing, let alone allow someone to write a book about it.
And even then, the evidence of evolution in our game thanks to the numbers will be subtle rather than drastic – an inexact science that will never yield perfection.
Good luck to them. At least someone has come up with a new term. Adapt or die!
The Port Adelaide example is a fascinating one, and really highlights how rapidly the game changes year-to-year. Educating the public and the media becomes even more important so we don’t keep trotting out Moneyball articles for another twenty years.
So What’s Next?
Where do we look for sport's next great data revolution? And what might it look like in the AFL?
For baseball, it seems to be a new arms race in player development and technology, dubbed The Wild West, where teams are exploiting an unregulated space in an attempt to improve their players.
Another angle addresses a critical aspect of organisational culture, combining key data with an unwavering commitment to the right people and messaging. Ironically this fascinating book, The MVP Machine, focuses heavily on the Houston Astros, who may have taken things a little too far and basically cheated during the 2017 season, aided by the illegal use of technology.
For football, there is a school of thought that the next evolution (or revolution) might have nothing to do with data or technology at all. It's about connection, about unity and brotherhood. The fierce desire to bring pure enjoyment back to playing footy together as a team, in an increasingly negative world.
And it's already started. The Tigers seem to have nailed it, but they won't be the only ones. If this semi-spiritual kind of movement gathers momentum, then a story like Moneyball will be just a distant memory.
Data or No Data, The Coach is still The Coach
In Michael Lewis’ follow-up book, The Undoing Project, he touches on the constant internal struggles teams face when applying new vs old methods of thinking.
“The enthusiasm for replacing old-school expertise with new-school data analysis was often shallow. When the data-driven approach to high-stakes decision making did not lead to immediate results – and occasionally even when it did – it was open to attack in a way that the old approach was not.”
The challenge for any team trying to fully embrace a new era of analytics (or anything, for that matter) might simply be to stay the course, finding the right way to merge the old and new together.
The challenge for everyone else is to move with the times and dig a little deeper. How many in our industry have read The Undoing Project? Probably not many, which is why we’re still talking about Moneyball.
A few years ago, Forbes magazine looked briefly at The Future Of Sports Analytics. It was called Moneyball 2.0.
Towards the end, an important question was raised. One that definitely applies to AFL footy.
“When does the coach get to be the coach?”
The answer? “Always”
In footy, the role of the coach is as important as ever. The numbers will provide just a piece of the puzzle, rather than solve it.
Coming Soon – Part Two: The Role of the Armchair Analyst