By Ted Hopkins
Stemming initially from University research centres, IT houses, business conferences and webinars, the name Big Data has started bobbing up and is spreading fast.
Typical is a recently glossy business magazine, Momentum, produced by the University Of Queensland that is headlined: ‘Big Data, Get it and get it right.’
How bizarre!
Since much of it is huge, difficult to grasp, and then assess.
Getting it is one thing, unravelling it is another.
I should know.
As early as 1995 I saw some of this phenomenon coming.
Instead of Big Data I called it Champion Data, the name of the small sport statistics company specialising AFL that I found and directed, from back then until October 2009.
The enterprise and my involvement subsequently triggered The Stats Revolution, which is the title of the book I have written on the subject, published by Slattery Media.
So much has changed in that period of almost 20 years—the humble statistic is no longer the humble statistic.
A vast difference now exists between Bruce McAvaney quoting archival statistics as embellished punctuation; compared to how numbers are collected, read, interpreted, and presented today.
In many respects, the numbers gathered now by armies of well-trained statisticians, their numbers available in real time, have now become the game and its language.
Players and teams are subjected to constant measurement by coaches, commentators, policy makers, consumers, and more recently fantasy football enthusiasts and betting interests.
The story of how statistics have infiltrated and evolved in AFL, their usefulness and abuse, is similar to other fields of endeavour.
Big Data is going to get bigger, more persuasive, and will not go away. At times it can be fruitful, and often dangerous. In between, there is lots of superfluous and meaningless junk.
To get the picture, let’s start with coaching an AFL team.
There are around 200 full-time coaching professionals spread around the 19 clubs. Add sports scientists, recruiters, and other specialists and the count is in excess of 500 on football department payrolls.
Game day commitments, recovery sessions, RDO’s and ensuring players are not subject to repetitive strain injuries means the actual time spent on the training track is limited.
What happens during the rest of the week?
Play table tennis? Eat lots of toast and jam?
Perhaps!
Mostly, the in-between time is spent sitting in front of computer screens poring over an ever-expanding array of statistics and multiple vision sources, or attending meetings in which PowerPoint has taken over as the Head Coach.
Legions of football coaches, commentators and their respective support staff have now become ‘Professional Analysts’ searching spreadsheets for the ‘data nuggets’ that will confirm their particular point-of-view.
There are even desktop applications that can do it ‘easily’ for them. Suddenly, coaches, commentators and officials can claim the title of ‘data savvy.’ Correlation and data mining are now lingua franca.
But this is just one way of looking at the Big Data question, albeit in its most obvious form.
Like an iceberg, there’s another thing happening under the surface.
It involves teams of highly skilled mathematicians, statisticians, code-cutters, visualisers, interpreters and their respective managers interrogating vast amounts of data.
Their collective priority is listening to what the data says.
Opinion is set aside because it can prove a distraction on the path of discovery. It is far too easy to mistake correlation with causation effects and to find misleading patterns in the data.
The sporting field and databases are rife with imperfections. Error and chance are also vital players.
In this alternative approach to Big Data, knowing the error rate is essential before any declaration of certainty is possible.
AFL season 2013 and the Grand Final provide a choice example of the differences between data used for spruiking or for knowledge.
From season start until the grand final a chorus of coaches and commentators declared ‘contested footy’ was the most critical factor for winning games.
Accordingly, it seemed players willing to use their heads as battering rams, became the way to ultimate success.
However, those who had been ‘listening to the data’ for discovery purposes knew otherwise.
Winning the contested footy count was obviously an advantage, but historically there were several other measures that rate significantly higher.
For example, kicking long to advantage proved consistently highest on the winning radar and poor kicking in the backline the worst thing.
As legendary coach Allan Jeans famously observed long before the advent of computers and Big Data, “there is no point winning the ball unless it is put it to good effect.”
In the grand final Sydney won the long kicks-to-advantage 78 to Hawthorn 53 and lost the contested footy count 140 to Hawthorn 170. They won the flag by 10 points.
How bizarre!
Big Data was right and wrong at the same time, depending on who was listening and who was spruiking.
Ted Hopkins is a Carlton premiership player and founder of Champion Data. His latest enterprize is TedSport, delving into the secrets of Big Data.