Skip to content

4 Things Stopping Analytics from Providing AFL Club Success (Part 1) A Look into the Role of Analytics in the AFL

Posted in Analytics Landscape, and Opinion

This post is dedicated to my musings about the footy industry and the role of analytics both now and in the future. There won't be much in the way of numbers or match analysis. So if opinion pieces aren't your thing, feel free to tune out.

There was some lively discussion on Twitter last night following this tweet:

I initially took this to be referring to the lack of a strong analytics culture among fans.1 I soon realised that Nick was talking more about the lack of analytics culture inside footy clubs themselves. At it's core, I think the two of these are closely related. Each one has the ability and potential to drive the other. But each also has different challenges that are holding it back. It's worthwhile to separate these two and look at their culture a bit more closely. Today I'm going to look at "analytics culture" at a club level. In a follow-up article, I'll step back and look at how analytics is embraced by fans.

Why Do Analytics Not Play a Bigger Role in the Club?


1. Hiring Skilled Professionals is Expensive Business

To be an effective head analyst at a footy club you would need to have a lot of skills. At a minimum, you would need to have knowledge and a degree in some sort of analytical discipline (probably Maths/Stats but maybe Finance/Actuarial Science/Other Sciences), be proficient in using and maintaining a relational database (most likely SQL), have fantastic statistical programming ability (probably at least R, Excel2 and whatever proprietary software the club may already use), be able to effectively visualise and communicate your ideas to those that don't understand stats (probably proficient in Tableau plus developed teaching/training skills), at least a basic understanding of fitness methods, injury and sports science, as well as an intimate understanding of the sport and it's surrounding landscape including historical and current day trends. On top of this you would need to be the sort of person that can manage an important role in a business by yourself with little to no help from any others (so probably significant past experience).

People with this sort of skill set are few and far between and usually rewarded with 6+ figure income from the private sector. Some may work for less out of a love for the sport, but it's not going to be for a pittance.

With a hard cap on footy department spending looking likely, finding money for an analyst will involve cutting money from other areas. A tough sell for any club.

AFL hasn't seen the analytics boom in the same way that bigger sports like baseball, basketball and soccer have, partly because the market is smaller. Finding a spare $100k is harder for an AFL club than it is for the New York Yankees.

2. Some Clubs Think They're Already "Doing Analytics"

This reply from InsightLane hits on a very good point:

The word "analytics" has been brandished with the "boring" tag of maths. Let me make this clear right here, "analytics" does not just mean using anything with numbers in it. When I use the word "analytics", it's shorthand for "looking really closely at something to try and understand more about it".

When I say clubs could benefit by using analytics, I don't mean we should whip out a big spreadsheet and all sit around it debating the difference between a player with 2.4 clearances per game and a player with 2.6.

What I really mean is that we should look very closely at each game and each player, watch the moments that change a game, study the difference between how a player functions in the forward line compared to when they play on the ball, understand which set plays bring more success than others and why they work. Doing this for your team *might* be possible by just reviewing the tape. But you need to do this for every single team and every single player in the league to gain a tactical advantage.

That's roughly 35 times the amount of men in this photo.

This is where the numbers come in. We just don't have the time, or in some cases the ability, to watch every single one of these things happen in real time and remember how they affected every other single moment in the game. But if we have the data, we don't need to. All the remembering and close watching is done for us.

A good analyst can find real patterns and plays in both your opponents and your own team, that you would have otherwise missed. They can also help you identify which of the patterns you are sure are there, don't really exist. This is where good "analytics" gives you a competitive advantage.

Clubs may think they're using analytics to make decisions on players and teams but it's far more likely that they are using "numbers" like KPI's and player ratings without really understanding what these mean in the context of a game.

Think of some of the big tactical experiments we have seen this year. We've seen clubs play without ruckmen, we've seen Essendon play all season without the majority of their first team, we've seen the highest scoring team in the league, Adelaide, play almost every game with an extra man in defence.

Is it the Crow's talented forwards or long-bombing backs that make them so dangerous?

Analytics can help us figure out whether Adelaide are high scoring because of the individual skill of their forward line, or whether they are high scoring because of tactical decisions that happen much further down the ground. Once we know this, we can go about stopping it.

3. Changing the Culture of a Footy Club is Tough Barrier to Break

AFL coaching staff, as a group, mostly consist of ex-players. This makes sense. Who else could have a better knowledge of the game than a person who has spent their whole life fully immersed in it? Nobody.3

The problem is, that by being on an AFL list from age 18, you often have to forgo higher education. While others are training the analytical, tactical and problem solving parts of their brain at university and in the workforce, an AFL player is by definition, training other parts of their body and mind.

A game of footy is a far more complex tactical battle than any game of chess could ever be. It only makes sense to have people who have trained in tactics, logic and game theory as part of your team making tactical decisions.

However, it's understandably very difficult for a senior coach (an ex-player himself) to give up any level of control to an outsider. It requires a level of trust in a person and idea that he himself knows rather little about. Communication and education is the key to developing this trust. But that is easier said than done.

4. There are Clubs Branching into Analytics, and They are the Ones Having Success

Okay, I lied. This one is not really a reason why analytics are not used, but a hopeful look to the future. As far as I know, currently there are no full-time employees working for any club in a primarily stats/analytics role, 4 but that doesn't mean that some clubs aren't experimenting with analytics and opening up their culture to change.

Clarkson has certainly led a new revolution in coaching.

Alastair Clarkson at Hawthorn seems to be a proponent of the scientific method in footy, and I personally know of researchers and analysts that have done work with both Hawthorn and the Western Bulldogs. I also know that the Senior Football Analyst at Adelaide maintains a fairly active presence in the online analysis community.

So, the three clubs that I know are at least opening their mind to advanced analytics include; Hawthorn - possibly the most successful team of all time, the Dogs - a plucky young side that is far exceeding most people's expectations, and Adelaide - a club that has had to cope with more setbacks in the last 4 years than most have in decades, yet still remain a very good chance of winning it all in the first week of October.

Whether these clubs are benefiting from a change of culture or this just a coincidence remains to be seen.5 But as I see it, there are some promising signs that the AFL may be starting to embrace analytics, although we still have a long way to go and many hurdles to jump.

 

In my next article I have a look at the analytics culture outside of the club rooms, among the fans, the TV stations and the online "fanalysts".


 

  1. Perhaps I have been spending too much time following the European Football "fanalytics" community who are currently going through somewhat of a culture crisis, finding any constructive criticism about people's work quickly devolves into shit-slinging.
  2. shudder
  3. This is not meant to read sarcastically. To suggest that anybody from the outside has a more intricate knowledge of the game than ex-players and long-time coaches would be laughable.
  4. Turns out I do not know much. Most, if not all clubs have full-time analytics staff. Whether these are strategic analysts or just glorified tech support probably varies from club to club. See the comments below.
  5. We'd need some sort of long-term analytics project to find out!

6 Comments

  1. Bill James
    Bill James

    There are several very true points within this article, the first being the ability of an analyst to question the view of coaches who have been involved in footy for their entire lives. This aspect however is very dependent on the coach, as some are receptive and inquisitive into the insights that good analysis can provide, while others believe their experience is enough for them to be able to coach the game effectively. Secondly it is also true that some (possibly a majority of) coaches believe analytics is just the use of KPI’s and simple numbers and then dismiss “stats” as not providing the full picture. This only changes as people with the skills sets mentioned (R, excel, SQL etc.) along with the communication abilities required to transfer the information within the footy field are hired into analytics type roles. This, believe it or not is actually happening.

    As a full time analyst in a club I do not know of a single club that does not employ at least one full time analyst. While perhaps not possessing the complete skill set described, most do have some grasp on these skills and are at least aware of the need to view the game on a more micro scale than broad KPIs and ranking/ratings provide. Like every other elite sport, footy is a competitive industry. If a club is seen to be gaining an advantage from a particular area, you can bet every other club will follow suit as quickly as possible. My club is not one of those mentioned as being at the forefront of analytics and we definitely aren’t, but everything I do in my role uses some form of analysis.

    The competitive advantage that good analytics provides only occurs when the insight gained is actionable. From a coaches perspective, that means something that they can actually use to improve player performance. Not all patterns that exist within the game are changeable and some that are changeable just don’t matter. Video analysis is particularly important in the education of players. Simple cutting and counting from video is becoming redundant however as player tracking data becomes ubiquitous. However the visualization of patterns within a game, identified via analytics, and therefore the education of players is made much easier through the use of video and the associated editing software.

    This blog and others like it amongst the “fanalyst” community demonstrate that despite a lack of data provided by Champion Data and the AFL, innovative analysis is still possible. It may seem unlikely but most of the Champion Data information that is not released to the public is not actually that useful. I will say though that I generally cringe when I see the “analysis” provided by the mainstream media, generally it sets the field of analytics within footy back to have some of that rubbish out there. The only consolation is that the coaches don’t pay any attention to it either.

    Looking forward to part 2.

    August 17, 2016
    |Reply
    • Thank you for taking the time to post such detailed and informative reply.

      Firstly I should say I'm certainly an outsider looking in when it comes to this topic. I know very little about the day to day operation of clubs. This article should really be taken as little more than an opinion piece.

      I did not mean to suggest that their are clubs that don't engage in any analysis, but I admit that it reads that way. The "no full-time analyst" thing had been bandied around a bit and I should have fact checked it before chucking it in. I'll edit this section to clear things up a bit and reflect what you (and others) have said, that every club does have *some* analytics team. Out of interest, does your role involve tactical insights and analysis or do you do more work at the player level?(eg elevations/droppings/trades)

      This article was supposed to be more about a general culture of flexibility from old ways and an openess to the scientific method. As an observer, I get a feeling of this from certain clubs more than I do others. The three I mentioned in the article stood out because people of prominence in the Twitter community (all of which follow mine and other amateur work) have worked with them before and I have heard good things. Some Bulldogs players even came to the recent MathSport conference in Melbourne. This is the sort of burgeoning analytics culture I was trying to highlight.

      I agree entirely with your point about actionable insights. Communication is the key to this all. You could have a 100 man analytics team solving the game from ground up, but if they can't communicate their insights to the coaches and players, this is useless. A similar thing happens with analysis in the footy media. If you can't get the point across quickly and clearly, you have failed. This is why I spend so much time trying to perfect visualisations. We need to be able to communicate analytical insight without having to train everybody up in analytics first.

      I agree that a lot of the CD stuff is not exactly game changing. But I firmly believe that if you give a team of smart people a bunch of player tracking positional data and time to work through it, they can come up with some pretty impressive insights and tactics for an invasion game like Aussie Rules.

      August 17, 2016
      |Reply
      • I'll admit to being one of the people who throw around the "no full-time analysts" statements. Like you say, it's untrue, but I should make myself clear what I mean specifically when I say it.

        As far as I'm aware, and the reason I put forward the poll question in the first place, is that no club hires has a full time analyst for the primary purpose of helping with list management and overarching tactical decisions and trends both for the team and in response to opposition analysts. It's mainly for other purposes

        Obviously I can't speak for 'Bill James'' or his club in question, but I believe (and may not be entirely true) that many of the clubs who hire analysts and those with IT proficiency do so in roles where that proficiency is required in sports science (physiology), such as GPS tracking data, 'helping' by being the go-to stats man for draft recruiters, or otherwise being the "data man" in relation to being a tool that helps the coaching and development staff.

        I obviously have no idea at the day to day roles of these analysts in question, but, for example take Adelaide's staff list:
        http://www.afc.com.au/club/information/staff
        Their analysts are listed under the coaching section, and three different analysts are applied to the three groupings of defence, midfield and forward lines. I would make an educated guess that the listing of these analysts suggests to me that these staff who look at data and then use this data to work in tandem with the development coaches to help develop these players, and work with the line/assistant coaches to help the individual player's skills and adherence to the tactical system. I'm not denying that these people aren't analysts in the sense that they analyse data and apply it to improve the team. It's just that they're not the type of analysts who work in the scope about answering questions how the game is played as a whole (like Darren's work about how the centre bounces are generally entirely down to luck) or work primarily in list management.

        Take an analyst at Fremantle who might look to recruit Brad Hill. He's a difficult player to gauge and value, because he's got an extremely low contested possession rate however has a very high retention rate kicking the ball inside 50.

        An analyst who might be hired to work entirely for list management purposes (which I'm not aware of any club has) as one person among many the list management team would try to statistically value Hill as a player by utilising statistical methods to remove his production from the tactical systems of Hawthorn, and then apply this new valuation of Hill as a player to Fremantle and their tactical systems. This then would be compared relative to the opportunity cost or a replacement player at Freo (or alternate list management direction). Of course, as you say, communicating this compared to "footy people" at other roles is the difficult thing.

        To make a further point about player tracking data and how it might make tracking footage redundant - that suggests to me that 'Bill James'' club sees player tracking data as a way to use data to help the development of a player much like video footage currently does. Video footage reviews with players currently helps their development like "okay look at your positioning at this stoppage. Look at your decision making with this kick. We can improve on that" and the comment above suggests to me that he sees player tracking data playing a similar role. But for example, I see player tracking data using a different role - a smart analyst in a club can use it, for example, to analyse the concept of "gravity" in ball movement and then determine the value of a big key forward to is like a magnet to high quality ball movement aimed at the top of the goal square.

        Just to chuck an example out there, hypothetically say player tracking data existed a few years ago. when the Dogs recruited Tom Boyd and everybody slammed his poor performances, the Dogs internally could have been confident because they could have used player tracking/gravity data to justify paying him $1 million because he's a very high gravity player, his mere height of 201cm players magnetising teammate's ball movement to be delivered in high shot location areas, which would not otherwise happen with a smaller key forward target. They could then use this insights in analytics to justify their pay packet in recruiting Tom Boyd above and beyond conventional statistics. But it seems to me that currently clubs see the value in player tracking as merely a current-players-on-list-development-and-coaching perspective, not to make wider, overarching investigations into revolutionary ways to analyse the game and make subsequent list management decisions.

        Of course, like said earlier, many coaches apply this sound logic intuitively or through other methods of application of hypothesis. The difference is that data can be more objective, collect 'big data' for every game and every player in the league, and can be another tool on top of people in list management and tactical roles in a team for what would be a big competitive advantage for a relatively small cost.

        But as far as I'm aware, there's no full time professional at any club with an analyst role that isn't somewhat related to sports science, individual player development/coaching or some other part of analyst rather than the roles that I describe. And when I mean "no full-time analysts" I mean it with people being tasked with those roles.

        I hope that all made sense. I'm happy to clarify anything here or on Twitter if asked!

        August 19, 2016
        |Reply
  2. Bill James
    Bill James

    My role branches into both tactical and player level analysis, mainly because it is difficult to have one without the other. For instance the desire to implement a certain style of play requires both tactical optimisation and the ability to identify individuals who have the qualities required to play that style. I continue to draw inspiration and generate ideas from the work of analytical “amateurs” within the football and wider sport community.

    The last point made about the player tracking positional data is well made. With the AFL/Champion Data taking control of tracking data from next season and potentially releasing the raw data to clubs there is certainly great potential for tactical innovation as a result. Unfortunately for analytically minded fans clubs will guard the most useful insights they get from the data, as well they should to not lose any competitive edge they gain. Instead I suspect fans will be inundated with basic distance/speed information that can already be found to a degree if you are prepared to read through various sports science journals. Perhaps though, some of the findings from this type of previous research will be overturned given a much larger dataset.

    Please don’t take this next point as a criticism it is more an observation. In other sporting fields it has been noted that researchers often confirm what the sports coaches already knew. This is because coaches in general, not just football coaches, are typically exponents of the scientific method whether they realise it or not. They develop a hypothesis about the way the game should be played and test it repeatedly. If their tests prove successful they retain the belief, if they are unsuccessful they either alter their hypothesis or lose their job. They are unlikely to have heard of Bayes or his theorem but coaches are also quite Bayesian in their approach.

    August 18, 2016
    |Reply
    • I agree with this last point 100%. In fact, I tried to make a similar argument in the article.

      The analysis process is already there. All big data and "analytics" does is provide a way to discover ideas without having to see examples yourself and rely on your own faulty memory.

      It also provides a way to test ideas and identity what is and isn't working much faster than we could without a rigorous system in place.

      August 18, 2016
      |Reply
  3. Bill James
    Bill James

    Nick,
    Not entirely sure how you got that I only see player tracking data as a player development tool, the thoughts I have surrounding the data generated by releasing all players tracking data are much further reaching and probably still are only a fraction of the tip of the iceberg. I was merely highlighting the fact that the "big data" produced will reduce the reliance on video analysis for clubs that embrace it.

    Not seeing the data being used in player development highlights a blindspot in my opinion. It's all very well using analytics to bring players into a club, not using analytics within the development and education of players leaves a significant amount of value sitting on the table.

    In regards to CD if you'd like to explain exactly what (please be specific) you believe you are missing out on I'm happy to let you know whether or not it is in fact available. I know having been outside the system previously that I thought I'd have access to a bunch of innovative stuff from CD once inside. Personally I use things that are publicly available far more frequently than the "secret sauce". This may be different for others, just giving my perspective.

    Anyway it's time to watch some footy.

    August 19, 2016
    |Reply

Leave a Reply