On Wednesday night we were lucky enough to witness a grown man – a professional football player no less – wearing a bra.
Of course, the bra was not actually a bra. The bra was a piece of kit designed to record just about every measurable known to man about Trent Hodgkinson’s physical performance.
Data captured by the bra is used to try and improve any future performances. This is sports science.
Sports science has become murky water to swim in following the ‘Blackest Day in Australian sport’ announcement last year.
However, another type of science involving fewer peptides is beginning to take a foothold in professional sports.
Sports analytics gained mainstream recognition when Brad Pitt became Billy Beane in Moneyball, with Hollywood shining a spotlight on sabermetrics and the maths behind winning at sport.
For those who haven’t seen Moneyball, it’s about baseball club finds way to succeed against rich baseball clubs by identifying a specific skill that is undervalued – how often a batter gets to first base – and choosing players based on that skill.
The key word in the sports analytics field is ‘undervalued’. As an organisation, it’s pointless to exploit something that is valued by everyone.
Baseball is unique because it’s extremely quantifiable. Every at-bat in baseball is a sentence within the larger chapter of a game. Each sentence can be broken down to reveal its minutiae with relative ease.
Analytics is now developing within more complex sports. The NBA is currently deep into the playoffs and basketball is having its own spotlight moment, with analytics theorists trying to identify and unlock the secrets to success.
Basketball is fluid, messy and complex. Unlike baseball, there aren’t clear-cut stop-start pieces of the game that can be broken down and mined for data easily.
In order to understand what elements of the game are undervalued when it comes to scoring the most points, analysts have had to develop new things to measure.
Of all the new metrics that analysts have come up with to try and understand what gives a team the best chance to win, the most interesting is Expected Possession Value or EPV.
EPV was created by two smart kids from Harvard. It’s designed to give an indication of how many points are likely to be scored at any given point in time during a game of basketball, and it’s incredible.
EPV uses ridiculous amounts of data about individual players with the ball (scoring/passing/assist efficiency), teammates on the court (same deal), and where everyone is positioned on the court (technology is awesome). It requires a Harvard supercomputer to then calculate all these variables into a number between 0 and 3 – the EPV.
I’m no Harvard supercomputer, but basically it works like this. Your team just inbounded the ball to a point guard on your own baseline? EPV is pretty close to zero right now.
Your team just inbounded the ball to an open Lebron James standing under the bucket? That EPV is as close to 2 as it can get. (You can read a more thorough explanation of EPV over at Grantland.)
In theory, if trends in EPV are monitored for long enough, teams work out the most effective way of scoring and defending. And winning.
EPV is far from perfected. At this point it’s better to think of it as a conceptual tool rather than a statistical certainty, but concepts like this will be the future. And in the future of sports, knowledge will be power.
In the US, there’s an analytics ‘community’ that share ideas and theories. There are annual sports analytics conferences. It’s normal for professional sporting teams to have people whose job is analytics. If you’re not trying to find the next piece of undervalued information that will help you win, you’re being left behind.
So are we missing a trick in Australian sports?
Basketball is one thing, but would it even be possible to quantify a player’s EPV at any given point within a game of Australian Rules football or rugby league?
Maybe it’s time sporting teams in Australia stopped putting bras on players, stopped looking at outdated Supercoach stats and started searching for something new and undervalued instead.
Karl Jackson (Champion Data)
Guest
AFL clubs are doing more than what is seen in the media. All clubs hire full-time analysts and some have six or more in this department. Further to Rourke's comment above, the research in the attached paper was developed into the Official AFL Player Ratings (www.afl.com.au/stats/player-ratings/ratings-explained). As Rourke said, research into this started in 2006. The ratings launched in early 2013 (well before EPV went public). In reply to Mister Football's comments about kicks, AFL kicks are broken down into distance, direction, pressure level, intended target and location on the ground. This gives us nearly 700 classifications for a kick, which are used to get a measure of difficulty for each attempted kick to then assess a player's kicking ability in a more accurate way than the simple kicking efficiency. This has been done since 2010.
Lukas Raschilla
Roar Rookie
Couldn't agree more, with new technology, we should embrace it. While sometimes it become overwhelming, such as stats like WAR (Wins above replacement) in the MLB. Efficiency ratings in basketball, QBR in the NFL, plus/minus in NHL and EPV is a no brainer and should be used in a variety of sports. Embrace the technology and analytics. I'm sure a number of clubs could benefit from this. Every aspect of the AFL (and NRL) has improved with increased knowledge, nutrition, player fitness, conditioning, and recovery so why analytics aren't being used more as in the US pro (and college) sports is baffling.
Rourke
Guest
The equivalent of EPV has been used in AFL since 2006. I've heard it being called 'equity', like the equity you've built up in your home - you build up scoreboard potential by getting quality possessions in good positions then hopefully realise it on the scoreboard. See http://www.rankingsoftware.com/research/PossessionVersusPosition.pdf
Knightblues
Roar Guru
I hate the bra, no I loath it, it makes the players look very unmanly for what exactly? Piss it off.
Vivalasvegan
Guest
Ahh... The romance of sport. This is why we all love it.
ChubbzyK96
Guest
Kick chase ability needs to seriously investigated. Check the enthusiasm levels and see who is chasing the kick and also getting back onside after breaks. If you have a forward who does this he is vital in D even if he isnt making the tackles.
Dan
Guest
run hard tackle hard
Brett
Guest
I believe the penrith panthers are practicing the moneyball theory in finding success in undervalued players. -- Comment from The Roar's iPhone app.
Peter Rabbit
Guest
One thing it failed to record : there's no G in Hodkinson.
Mister Football
Roar Guru
Here's a site that puts the spotlight on the use of R (an open source statistical analysis tool), in particular, the use of big data and predictive analysis in sport. http://blog.revolutionanalytics.com/sports-1/ As it happens, towards the end of the page, there's a small article on the AFL, collecting various stats on "kicks" and being able to predict a winner from this. The AFL is actually quite suited to this sort of analysis because literally a thousand different statistics are collected each game, and some of those statistics are predictors of winning games of footy. Most clubs are already investing heavily in video and computer-aided analytical systems (in fact, I'm sure I read somewhere recently that an AFL club had sold a system overseas). Over the past decade, we've learned a lot more about being able to split the key stats into lower levels of granularity, coming to understand that one possession is not equivalent to another possession and one kick is not equivalent to another kick (there might be up to 10 sub-categories of "kicks", and in turn, each of these can be further sub-divided). As one small example, a contested mark is far more valuable than an uncontested mark (generally speaking), because it usually means you have beaten one or more opponents to a ball which was contestable and won clean possession to for your team, and it follows that a contested mark close to your own goal is even more valuable again because it leads to a direct score, etc. Then we have the split between winning contested footy and loose-ball gets, and we certainly value the former, but the latter can hurt you just as much, and if one player in particular is excelling in that stat, it means he has been allowed off the leash a bit too much, then the question is where is that player collecting the loose ball gets? are they leading to inside 50s? that's important because - refer to stat on marks inside 50s, etc, etc.
Swampy
Guest
If be extremely shocked if clubs as wealthy as collingwood didn't have someone looking at this area of development. There is a great debate in basketball going on between the pro-analytics and anti-analytics bases. It is still an experiment in progress. I believe the Houston Rockets theorise that all shots should be taken either as a three point attempt or from within 10' of the hoop with no shots taken in-between. As such they have run their aligned D league team as an experiment based on this approach with great success. However, the theory was a monumental failure in the recent playoff series against Portland - largely resulting in their first round exit from the playoffs. Back to the drawing board. -- Comment from The Roar's iPhone app.