How do we understand the value of Key Forwards?

One of my biggest issues with the current state of how footy analytics is developing is that it’s oriented toward player production metrics, rather than player value metrics.

What do I mean by this?  Well, the development of analytics in the statistical value is geared to much toward understanding the value of a player’s actions on the ball on the field, but not in the wider context of the process toward winning or the wholesome value of a player directly and indirectly.  SuperCoach/Champion Data Rankings Points and AFL Player Ratings points, whilst effective if you wanted to ask yourself “how good was this player’s statistical game” don’t answer the inherent question of how that improved that player’s team’s chances of winning.  It measures the production, on “ball-involved” or “on-ball” aspect of a player’s performance, but nothing more than that.

For example, for defenders, it’s hard to statistically analyse how they helped their team to win.  Intuitively, a keen footy eye will realise that a defender is a good player because of how they adhere to successful structures and as such make opposition ball movement poorer.  But it’s difficult to analyse in a statistical sense.  For example, if the coaching staff have implemented a successful defensive structure, such as packing the corridor for central throw-ins because the opposition moves the ball through the outside following the throw-in, a player’s value is how apt they are at following that structure and blocking the corridor to force the opposition into a more unusual corridor ball movement.  But valuing that analytically or statistically?  How do we give a numerical defensive value to the player who does that well?  It’s tough.  Conceptually, defensively, is all about preventing the opposition from doing what they usually do so you’re measuring the success through the lack of something, rather than the value of actually doing something which is easier to measure.  It’s not a concept that’s unique to footy – for example, it exists in all sports like Basketball (long live Grantland).

But today I’m going to look beyond measuring defence where it’s hard to do so analytically, and I’m going to talk about another idiosyncracy of footy as we know it.

Australian football does not have an offside rule.  It’s been a feature of the game for over 150 years, as old as the sport itself!

As such, when you’re wholly talking about footy analytics, I don’t think that you can categorise it into what some sports analysts have dubbed “territory-invasion” sports such as Rugby and Soccer, this being one example.  “Territory” is a more distinct concept in those sports, because of the nature of an offside line and the protection of territory behind the offside line.  Not saying it doesn’t exist in Australian football – your defensive 50 is certainly “your” territory, but the concept of “invading” territory behind ball or player was smashed not long after the game was invented over 150 years ago… by Tom Wills, the father of Australian football.  How he did it is explained very well in Time and Space by James Coventry, so if you haven’t already, picked up a copy and read it yourself.

If you’ve read my stuff, I go off a tangent often… but it’s all linked back to what I’ll write from here on it.  Because the lessons we’ve learned in analytically analysing defenders we can also apply to other positions, most pertinently key forwards and their value … and it’s not something that we can learn from other sports, because players ahead of the ball given no existence of an offside line impacts the quality of ball movement, and that gives values to key forwards, as these key forwards typically play in front of the ball and impact ball movement.

If defensive roles have value because a defender’s action and inaction changes the production of the opposition players for the worse, such as worse decision making in ball movement because of good defensive positioning, that’s something we can also apply to forwards conceptually – even if it’s something that we can’t really statistically measure.  Key Forwards have value because they play in such a way that adds value because how make their teammates produce better actions, and their oppositions worse actions.  Okay, I’ve minced my words a bit so read that again if you have to!

If we compare salaries to analysing “production”, we all understand it for defence.   Nobody is suggesting that Alex Rance is overpaid, despite the fact he’s being paid somewhere in the vicinity of $600k a year, ie one of the highest paid players in the league, even though there’s an analytics disconnect between the intuitive (and correct) value of his salary.  I mean, he was rated roughly outside the league’s top 60 players according to AFL Player Ratings Points at the time he signed that deal, but we were able to “explain away” the disconnect between that and the contract he signed (and he could have fetched more from other clubs like Brisbane) because of the fact we all understand the difficulty in measuring defence analytically.

So what other position also have disconnected between what their value is measured by their salaries, and their analytical performance?  Key forwards, because the play ahead of the ball and impact other players on the field.  Cloke’s last contract was a lot at the Pies, Freo were throwing around millions at any key forward they could approach over the last two years, and obviously Tom Boyd signed a multi-million deal as a teenager.  There’s clearly a disconnect with how much they get paid, and conventional, analytical ways of measuring their on-ball production, as no key forward is in the top 16 of AFL Player Ratings Point.  But why is there a disconnect?  Because of the same reasons – their value isn’t so much their on-ball production that can be measured statistically, but how they change the actions of their teammates for the better, and their opposition for the worse.

A lot of the top draft picks and massive percentages of the salary cap goes to key forwards.  Sydney are literally paying around one-fifth of their salary cap for two players – Tippett and Franklin – and there’s a disconnect between how much they’re getting paid and their “analytics of direct production” given that Franklin is “only” the 17th ranked player currently in AFL Player Ratings Points and Tippett all the way down in 80th.  For defenders it’s easy to “explain away” this difference – we understand it conceptually, even if we find it difficult to measure it differently.  But for forwards – well, we need another left-field thought process to get there.

In case you’re wondering, yes, it’s thoughts like this that tick in my head when I can’t get to sleep!  Basically, conceptually, how can we “explain away” the difference in analytical production measurement and the (correct) intuitive value of these key forwards that is explained by player salaries?

My initial thoughts are this:  key forwards do two things that give them inherent value.  Firstly, they help their team’s quality of ball movement (ergo not so much through their own direct actions, but rather how they impact on other players on the field much like a defender would.  Except in this case it’s the teammates and not the opposition, and for the better and not for the worse).  The second thing they do is that they play a pseudo-defensive role in their opposition’s attack from defence.

How exactly do key forwards help their team’s ball movement?  Well, we have to think about a key forward vs a resting midfielder or a small forward or whatever the alternative to not having a key forward is.  Basically, picture a typical ball movement scenario – whether it be bursting from a stoppage, switching the play, quick ball movement through the corridor or whatever – but basically, picture those scenarios to be identical, but ahead of the ball (which they can be because there’s no such thing as an offside), the player is different, a tall player vs a small forward.  Does the quality of the ball movement change because of the difference in player ahead of the ball?

My answer is yes.  Firstly, most obviously, a smaller player is likely to be intercepted against than a key forward (“halving a contest” using the technical term).  But it goes a bit deeper than that.  I think the quality of ball movement is improved because a key forward is less likely to lead to the flanks or pockets than a smaller player because they can create their own space (pushing smaller players out of the way etc.) rather than having to find space near the flanks, and the central position of the ball movement is a better shot location than a pocket – for example, a goal is more likely to be kicked if the spill from the contest is directly in front of goal 30m out, rather than in the pocket 30m out.

But it’s not just that.  The presence of a big man up the ground helps ball movement through decision making.  Teams are more likely to simply move the ball closer to goal because they have the confidence that they have a big unit up there who might mark the ball.

Take the Western Bulldogs‘ loss in Round 6 this year to North Melbourne.  The Dogs lost Tom Boyd to injury the game before, and didn’t have a key forward in the team, with only a second ruck in Roughead or Campbell resting forward that match and the rest of the forward line being small or medium players (I still get nightmares of us kicking it to skinny Bailey Dale as a marking target … the nightmares are less scary with a premiership cup however!)  The quality of the Dogs‘ ball movement was poor, because they chose not to kick toward goal because of the lack of tall timber up forward.  Lachie Hunter, who played on the half back flank, won 44 touches because of the fact the Dogs over-used the ball sideways and backwards because of the lack of structure forward of the ball impacted their decision making and confidence in moving the ball in a direct-to-goal way.  The Dogs ball movement, which can be measured in the sense that they kicked just six goals for the match, was poor.

I’m hoping with player tracking data from next year, that’s the next step.  The Bulldogs have mooted a hackathon next year with this type of data, and if I had the computer-whiz skills to code this, the first thing I’d try and measure is the “player-gravity” nature on how some players impact the quality of ball movement ahead (or behind) of the ball.  All things being equal, do key forwards really get the ball to gravitate toward them in better ball movement locations, and if so, which key forwards are better at doing that than others?  I’ve always said that I find it a bit strange that somebody as big as Tom Hawkins leads to the flanks, rather than crashes packs in a more central location, which is obviously a better place for the ball to be inside 50 – but can we prove it analytically that Hawkins is a poor “gravity” player using player tracking and ball data?

The other reason I think key forwards have value intuitively, as measured by their salaries, that can’t be measured directly by ball-involved analytics, is that they play a pseudo-defensive role on the oppositions defence.  What do I mean by that?  Well basically, the opposition change their way they attack from the defence given the mere presence of a player in the forward line.  Simply put, if you’re a big unit, you’re not going to be left alone that often, because you don’t want a player zoning off then a different player who zones on being not key defender sized and being outmarked.  Simply because of the size of that player, it’s more destructive if a big guy gets zoned off than if a smaller player gets zoned off.  And that, even indirectly, impacts the quality of how an opposition attacks from its defence.  The inflexibility that the opposition has in defending a bigger player impacts in attack therefore logically it is a sort of pseudo-defensive value of a key position forward.

Picture this.  Imagine Tom Lynch in the goalsquare against, say, Heath Grundy.  You’d say that Grundy is a good chance of defending Lynch, but Jake Lloyd playing as a running defender isn’t.  But if Tom Lynch was replaced by a smaller player, say Aaron Hall as a small forward, you could play Lloyd on the smaller player, have him run off that small player with the knowledge that Grundy can zone off and pick up Hall if worst comes to worst, and it’s not the end of the world and in fact means that he’s likely to intercept against Hall.  But Lynch?  He impacts how Sydney would like to attack and position their defenders for optimal attack out of defence.

That’s what I love about the unique nature of the Australian game – it, as a result, gives unique concepts to think about in an analytical sense.  Trying to understand the value of key forwards is just one of them.

Advertisements

The four list management approaches AFL Women’s clubs are taking

If a friend and I were to pick schoolyard teams from all NBA players in history, to battle head-to-head, and I had the first pick, I wouldn’t pick Michael Jordan or Wilt Chamberlain.  I’d pick Dennis Rodman.

It’s counter-intuitive, but I think it’s the best pick that one can make.  It’s not because Rodman is the best player in history – in fact, he’s far from it – but it’s because it’s the best way to build a winning team.

Rodman was a much better rebounder and defender than the next best player in history, even if he wasn’t a primary scorer that made him productive, which makes him the best player to construct a team – even when accounting for the fact that rebounding is only about a quarter of the game in basketball, he was so much better than the next best rebounder in history than Jordan was better than the next best scorer, it means that my team is more likely to win, even with my friend picking the GOAT Jordan with pick 2.

Another way of thinking Rodman’s inherent value is how the game of basketball is constructed – only one player can have the ball and shoot.  So if every team has two primary ball handlers and scorers, who take up most of the possessions anyway, who do you want as your third-best player – someone who is a good scorer, but won’t get opportunities to shoot because he’s only the third best shooter, or the best rebounder because there’s always rebounding opportunities?  In other words, Rodman was the by and far the best 3rd best player on the floor of 5 players in an NBA team, relative to the other 3rd best players in NBA history, more-so than Jordan was a better primary scoring option than, say, Kobe Bryant was a primary scoring option.

Don’t worry, I’m going to segue this into footy.  It has got me thinking – if I were to do an all-time draft for footy, who would I pick first?  How is our game constructed in such a way?  Would I pick Wayne Carey, Ted Whitten or Leigh Matthews first?  One was the best key forward in history, one was the best player in history given his versatility, and one was a player with scarce production types, an elite goal-kicking midfielder – so how do decide between the three to increase your likelihood of winning?

If I were to do it for currently active male AFL players, I’d pick Alex Rance or Cyril Rioli as the first player to pick – not that they’re directly the most productive players in the competition, like a Patrick Dangerfield, but they are player who rare in that they’re good in multiple categories – Rance is a lockdown defender, structural anchor, and rebounder all at the same time, and Rioli is an overhead, ground and pressure threat in the forward line all at the same time.

The issue with playing games like this is that it’s all hypothetical – there’s no way that, in reality, that the AFL will be blown up, all players redistributed we’ll be able to see a draft like this occur or to test my theory here that passing over Dangerfield for Rance for pick 1 in a “dispersal draft” would play out in the real world.  It happens in fantasy footy or in American sports when single folded franchises have dispersal drafts, but it’s not really the same thing, is it?

This is why, from a footballing perspective, the AFL Women’s competition is so fan-bloody-tastic.  It’s fan-bloody-tastic because socially it’s a provision for a fully national and professional competition for the female footballers who didn’t previously have the opportunity to play at this elite level, but that’s not what I’m writing about here.  It’s so fantastic because it’s an opportunity to see how these mad scientist theories about the game play out in reality.

The competition provides opportunities to analyse and compare the philosophical approaches to list management and to see how different clubs are building their lists, and to give opinion on which clubs are gaining talent in ‘marginal’ areas, where the drop off in positions, and how clubs can shape lists to build their tactics.

I’ve come up with four philosophies that I think the clubs are taking in building their list:

The Engine Room Philosophy:

One school of thought is that by having gun midfielders, you can influence the play.

It’s certainly not a bad philosophy.  Midfielders are around the ball more and can influence the play more often.  If you win the tough balls, tackle, harass and win the clearances, and have goal kicking midfielders, the game can certainly be broken open.

It raises two fundamental questions however – what is the marginal value of midfield talent, and what’s the drop-off from adding more midfielders?

The first part is, if you recruit a handful of the best midfielders in the competition, there’s still plenty of good midfielders left.  Teams such as the Western Bulldogs who went for Ellie Blackburn and Emma Kearney perhaps gave up the opportunity to recruit an outside player, forward, back or ruck, to cement the first opening bounce combination.  However, that still leaves midfield talent on the table, who might not be that much worse, such as Karen Paxman who signed as the final priority pick of any of the Melbourne-based clubs.

The other aspect is the marginal drop-off.  Does a player who is good in isolation have the same impact if she’s the first, second or third best midfielder?  Consider Melbourne with Daisy Pearce, by and far the best midfielder in Australia.  If she gets tagged and defensively minded, it might mean that the other midfield players for Melbourne play better with less attention.  Pearce and a player who was picked up in the second or third round of the draft could, as a duo, have the same midfield impact as a Kearney and Blackburn inside midfield duo – however Melbourne didn’t burn two of their first three signings on the midfield.

Clubs with the Engine Room philosophy: Western Bulldogs, GWS, Fremantle

Positional Versatility philosophy:

I love this theory.  In signing the marquees and priority players well before the draft, there’s still a lot of uncertainty as to how the draft will play out (unless, you know, a club’s draft board was to be published online property was to get out…).  The worst thing that could happen to a club is to approach the draft targeting a specific position, only to see those players go just before their picks, because the players that they’ve already signed doesn’t have any versatility to the positions that they can play.

For example, say the Dogs with three midfielders (if Katie Brennan plays more midfield than forward like she did in the All-Star game), are clearly not going to target a fourth inside midfielder with their first draft pick.  But what if the draft pans out in such a way that the best talent available just so happens to be inside midfielders – all the gun defenders or outside midfielders like Steph Chiocci or Hannah Scott go before the Dogs get an opportunity to select them, leaving them in picking up more inside midfielders even though they’re not the best inside midfielders at the club?

Carlton in my eyes took this approach – in signing Brianna Davey, Lauren Arnell and Darcy Vescio, they picked players who are flexible in their ability to play various positions giving them more leverage in the draft.

Davey is a key defender who is flexible enough to play as a lock-down player, rebounder or even stints through midfield.  Arnell is a mid-forward who is equally adept inside or outside in midfield, or as a small forward, and Vescio’s agility, pace, and goal-sense makes here capable as an under-sized full forward, as a small forward next to a spearhead (like she played alongside Moana Hope in the All-Star game) or even as a high half-forward or an outside midfielder.  Carlton can now approach the draft with flexibility, depending on how other clubs draft – if gun forward players are on the table, they can take them, knowing that they’ll move Vescio and Arnell to more of a midfield role.  If there’s still plenty of inside midfielders in the mid to late part of the draft, they’ll be able to keep them in the forward line.

However, I think that this philosophy falls down in two ways – because of the free agency period, and because of recruits from other sports.

Firstly, the free agency period allows clubs to sign three players after the draft is concluded.  This allows them to select unlisted players from any state (and those that are willing to relocate) – which means that clubs that the flexibility leverage benefit in the draft is limited if there’s a free-for-all anyway after the draft.  The Dogs, for example, could draft even more midfielders and sign key defenders or key forwards when the draft is all said and done – clubs are not going to be out of the lurch because of an imbalanced draft, because they can rectify a badly balanced list in the free agency period.

The signings from other clubs also diminish the benefit.  These athletes, from a range of sports stretching from Athletics, Basketball, Netball, Soccer and so on, are generally athletic, but relative to the rest of the competition, haven’t played a heap of footy.  That means their development can be shaped in such a way that they can play in a position of need, or they can be moulded in such a way that they’re a utility.  It’s similar principle to how Mark Blicavs has operated at Geelong – as a player who had been not been “cemented” in any given position when recruited, he’s played as a 1st, 2nd, 3rd ruck, key forward, key defender and big-bodied midfielder in the entirety of his career.  With every club having players from other sports, the same philosophy applies – these clubs are able to pivot around their recruiting by placing their other-sport poached players into positions of need.

Clubs with the Positional Versatility philosophy: Carlton, Adelaide.

Don’t put all your eggs in one basket philosophy:

This is also sound logic – if you sign your early marquees and priority picks across various positions, they make the whole of that position balanced.

For example, a gun key defender can anchor the structure of the defence and allow others to play less critical roles.  A good ruck can influence the stoppages and make weaker inside midfielders better.  A very good key forward can attract double-teams and allow teammates to get off the leash.

Collingwood, for example, are taking this strategy.  They’ve got the best forward in Moana Hope and the best ruck in Emma King, which means that their stoppage and forward line start off better than any other team in the competition.  They also have a mid-sized defender/midfielder in their operations manager Meg Hutchins signed, which again plays a different position to the other two.

This allows for there to be at least a base level of ability in all positional areas of the game – Collingwood will at the very least be competent up forward with Hope and at stoppages with King’s influence.  I’ll bet my bottom dollar that Collingwood will not be the among the worst teams in pure forward efficiency or clearance effectiveness.

The issue with this strategy is that the specifically good players can be limited without support.  Imagine if Collingwood took one three strategies to the draft: Improve forward line to support Hope, improve the midfield to benefit from King’s ruck dominance, or to have a balance between the two.

If either of the first two strategies are taken, Collingwood will have major problems – without a midfield to deliver the ball to Hope, she’ll get terrible delivery and defensive structures will be able to stop her.  If Collingwood’s forward line (apart from Hope herself) is the worst in the competition because of Collingwood not wanting to waste King’s stoppage ability, teams will be able to zone off, double-team and sit on Hope for the entire game and limit the space she has to work with or one-out opportunities against defenders.

At the end of the day, however, Collingwood do have the best forward and ruck in the competition – the match winners in the All-Star game – and surely that counts for something

Teams with the “don’t put all your eggs in the one basket” philosophy: Collingwood, Brisbane (to a smaller extent), Carlton (to a smaller extent)

Structure, structure, structure philosophy:

Teams that force other teams to worry about their tactics, structure and limit the effectiveness of their opposition’s strengths could be an effective way.

Imagine a team forward planning a meeting with Brisbane – “how the hell can we get our relatively weak defence to compete with the duo of Tayla Harris and Sabrina Frederik-Traub up forward”

Brisbane might not have picked up the gun inside midfielders, rucks or key defenders.  But by picking up the two best, tall overhead and contested marking threats in the country in Harris and Frederik-Traub, clubs are certainly going to be worried about their marking threat.  Get it to them enough times and they’ll convert.

In fact, Brisbane’s threat with tall forwards could be a bigger threat than any one team in any one position – no team’s inside midfielders, defensive unit, or forward unit, as a collective, is as good as Brisbane’s forward overhead marking threat.

That might result in opposition diverting resources into trying to stop them to their own detriment.

Melbourne are also doing this, by recruiting Mel Hickey in defence as a rebounding centre-half-back, as are Carlton with Davey.

Structure behind the ball defensively can influence your opposition’s decision making and limit their effectiveness.  Engine room midfields are no good if you’re turning it over to the defence because of better opposition structures.

In other words, it isn’t all about building the best pure talent or productive players, rather, giving the opposition plenty to worry about structurally due to being dominant behind or in front of the ball.

Clubs with the structure, structure, structure philosophy: Brisbane, Melbounre (to a lesser extent), Carlton (to a lesser extent)

It certainly will be interesting to see how clubs pivot around these initial philosophies in their future drafting and recruiting.

Also: my AFL Women’s predicted ladder:

Given the above philosophies and what I think will be the most effective, here are my thoughts on how the ladder might shape up:

  1. Carlton
  2. Collingwood
  3. Fremantle
  4. Western Bulldogs
  5. Melbourne
  6. Adelaide
  7. Brisbane
  8. GWS

 

A rant on my view of footy stats!

When I finished high school at the end of 2014, there was about six months of time to fill before the start my university life.  Though I worked, partied and played cricket through the summer and enjoyed turning 18 in that period of time, six months is a lot of hours to fill.

So to fill the void over that period of time, I watched NBA basketball – I mean an absolutely large amount of NBA basketball.  I had occasionally watched my team, the New Orleans Pelicans beforehand, but now with the power of NBA League Pass, I watched the vast majority, mostly live, of the Pelicans 86 games that 2014-15 season as well as the majority of other neutral blockbuster games that NBA season.

Whilst I’m primarily a sports fan, so I was amazed at the superhuman feats of one Anthony Davis, there was a further itch that just had to be scratched.  That itch was understanding the processes of NBA basketball – understanding how and why Anthony Davis was such a good player, what were the tactical systems toward winning how we can use numbers and data as a tool to understand this.  There was plenty of good information on the internet, but when the internet was not quite enough, I read the most brilliant book in all sports analytics, Basketball on Paper by Dean Oliver.

That book absolutely blew my mind – it gave new understanding to how the game of basketball can be segmented, broken down and understood in new ways using easy to comprehend statistical methods.  “Pure” player valuation, four factors of basketball, and the nature of breaking the game down into possessions to determine efficiencies were whilst completely new to me, were entirely rational and were data that helped me understand the processes as to how teams win, which was part of my interest in the NBA in the first place.

This then got me thinking about the greatest sport in the entire world – Aussie Rules Footy – and how come there is no similar analysis of the game in the same sense.

This is where the confusion came from.  I knew numbers existed in the AFL, but they were for entirely different purposes.  Journalists used numbers to further a narrative (whilst it wasn’t quite “data journalism”, a typical article would explain how a player played a good game with his disposal count which is entirely appropriate and okay).  Coaches used numbers for Key Performance Indicators or a short hand way to communicate with other coaches or their players.

However, none of this was what I was ultimately interested in: which was, using data and analysis help a team win and investigate the game – to come up with purer versions of player valuation, to break down the mechanisms in which teams score, and to support list management through the use of data.

The main issue here was, that, the use of such statistical data seemed entirely rational and a great way for a team to improve their chances of winning – after all, that’s what the whole book Moneyball is about – but it was of no more particular interest to me than other areas of footy.  I am footy mad, and I am interested in all areas of the game – administration, management, tactics, list management, the stories behind the scenes and absolutely everything to do with footy at a professional level.  I am interested in the tactics and list management and this is an input into teams ultimately trying to win.  However, this is done to death in the media and in clubs internally – the media oversaturates the coverage of trade week; and internally at clubs, they invest massively through having multiple full time scouts and recruiters.  There was no reason for me to gravitate toward contribution something that was done to death already, and people with greater credentials were talking and speaking about.

What I did gravitate toward was how data was not properly applied in understanding the processes of on-field action as a whole, like what was described in Moneyball and Basketball on Paper. I wrote a couple of confusing pieces about it and my tiny readership included some people within the industry, who welcomed a new viewpoint into the game.

In simple terms, if you ask yourself the question – “what are tools that can be used in order to help a professional AFL football team win”, almost everything is checked off by either clubs internally or the media.  The media and clubs spend plenty of time and money analysing list management, clubs invest large amounts in physical and sports science (to the extent of the Essendon doping scandal), the media is pretty terrible at video analysis but at least the clubs aren’t, there is plenty of stuff in both the media and clubs about club culture, administration and business, and so forth.

What I found that the footy industry as a whole, the media and clubs lacked was the use of data (apart from outside a very niche grouping of people which I communicate often with on twitter).  There was a lack of good data journalism in AFL football (that can be seen with the writing of Zach Lowe and FiveThirtyEight), and there was a lack of good data application to understanding the processes toward winning in clubs – MLB and NBA clubs had hired dozens data analysts to help their equivalent of tactics and list management, which did not exist in the AFL.  (There are data analysis employed in AFL clubs but they fill the roles of individual player development, video analysis, translating KPI’s, physical science and so forth, the pre-eminent data guru who works purely helping his team win tactically and breaking down an efficient use of resources backed by data is Darren O’Shaughnessey who only works as a consultant, not a full-time employee at Hawthorn).

The fact that there was such a big gaping hole was evidenced by a fact that someone unqualified and as young as me had the opportunity to fill a whole to think about it – at the time I was only 18 years old, had no coding or computer ability whatsoever beyond copying and pasting AFL Tables into Excel (and still don’t), did not do maths or statitsics beyond year 12 Maths Methods, had not done any journalist or web design training (hence the terrible quality of my writing on the terrible quality of the website) and was only beginning the first year of university.  I’ve never shied away from the above and fully admit that I am young without expertise in those areas!

What I was fuelled by, however, was a love of analysing the game analytically.

So then, in 2015, I started this blog, Madness of Sport.  It wasn’t meant to be anything impressive.  In fact, I consider it more of a diary or a portfolio on my wider thoughts about football, rather than any motive to send information out into the world or appeal to the masses.  I don’t spend any large amount of time editing what I have to say, I write when I feel like it, and I cover plenty of different angles – my thoughts as a Bulldogs fan (which I did a lot more of early on, simply reviewing the games that the Dogs played), my thoughts on analysis of the game, and my thoughts of the use of statistics in football.  In time, as I develop my business ability through my university education, I’m likely to blog about international sports administration.

I gravitated toward twitter and the other fanalyst websites.  Not all of them did exactly what I wanted out of football – websites like Matter of Stats does a lot of things, but mainly team ratings systems, whilst The Arc is graph/data journalism, whilst Hurling People Now uses stats in an intelligent way, but not to entirely break down the processes of how teams win – their use of hitout ratios, which is clearly debunked in terms of statistical significance toward the processes of winning, an example of that.  These people, whilst not quite breaking down the actual data processes that constitute winning, such as the breakdown of possession chains and the relative value of players to a replacement of marginal amount, shared something in common with me – they looked at the game through a rational lens, largely used numbers and data to support their assertions, and didn’t buy into the media analysis of the game.

So I went on my cheery way for well over a year, enjoying posting my insights into the game here on this blog and enjoying the little small community of people who rationally think about footy on twitter.  In this period of time I did as much research and independent thinking I could, like how AFL Player Ratings Points could be broken down more efficiently for a more context-sensitive player valuation system, and I lived the life of a normal young adult!

Over the last two years or so. I’ve developed a whole heap of independent thinking, such as the valuation of inside midfielders and the true importance of ball movement, but I haven’t posted everything due to time constraints, the lack of public data that’s available, the lack of multiple video angles that’s available, and the fact I want to keep some things under wraps should I ever be employed in the footy industry and I can offer people something that isn’t already public!

Last month, I split my two twitter feeds, sport and personal, aside, and I was a lot more aggressive with my new footy twitter feed.

Last week, I blogged about something that had been playing on my mind for a while – the true value of contested possessions. The funny thing is, what I’ve said shouldn’t be of any great revelation to an astute analyst of footy, and had time to think about the game critically and independently.  Due to the nature of it being 3500 words, I wasn’t trying it to be accessible – I didn’t do any promotion of it rather than just sharing it on my own twitter feeds.  In fact, it was simply like what all other blog posts here are – much like a diary entry on my thoughts on the game that I just felt a desire to cement somewhere else!

As such, the piece wasn’t so much to criticise any one use of contested possessions, or to criticise the media, rather than to just commentate how the understanding of contested possessions is driven by its use as a narrative device or KPI in clubs.  This is to explain how it not the same thing as analysing  numbers as part of the data processes in understanding how teams win.  I further explained the divisions of numbers in these formats yesterday.

What happened today, was interested as it was a further conflation of these uses of numbers in footy.  My piece was looking at the use of numbers in the game from an analytical sense, to understand the mechanisms in how teams win, much like what Moneyball and Basketball on Paper did all those years ago.  It wasn’t to offer a narrative as to how Hawthorn do or don’t win or to preview a game in a narrative sense – because people who have been doing it longer than I have been alive, like Rohan Connolly, can do so in a much more articulate and accessible way than I can.  They’re experienced, trained journalists, I am not.

Rohan Connolly today goes on radio without truly understanding this use of numbers in football.  He hadn’t read my post, which quite clearly demonstrates the analytical value of contested possessions and how the media narrative use of contested possessions confuses the issue, because they aren’t the same thing.  I have never claimed that Connolly is anti-stats, and I agree that he uses stats.  Rather, he’s not expert in using data to analyse the processes toward wins, and his distrust in that (and trust in things like gut feel) is an example of the archaic thinking in footy that I attempt to feel.

(And really, Rohan, three of the six teams left in the competition, Hawthorn, Adelaide and the Western Bulldogs, are three of the most prominent teams in using data analytics to guide their list management and tactical understanding of the game; and it was the “gut feel” and mistuse of KPI’s as analytics that sees teams like Collingwood slipping down the ladder)

 

He also didn’t quite understand InsightLane’s explaining of the use of contested possessions in analytics (perhaps it was too technical for him!):

That’s the problem.  Andy Maher admitted he got bored by numbers.  It’s Rohan Connolly’s job to talk to the masses – something that I have never claimed to do.  So their understanding of numbers in footy is different to mine and the consistent readers of this blog.

Connolly, I believe, is one of the good guys in footy.  He deals with his fair share of trolls on twitter, he has genuine enthusiasm for the game, and has promoted the game for over thirty years in a media role.  I will always attempt to support and get along with people who have genuine enthusiasm of footy as a whole.

The issue is here that he confused his use of numbers and people like me who use numbers in talking about how teams win, because I believe that there is a gaping hole for that in footy as a whole.

One day I hope to work in a footy department of a professional football club, using data, video and rational thinking to help that team win.  Until then, I blog and I tweet.

Numbers in footy: don’t confuse the uses!

A short, quickly written article about the role that numbers play in footy for anybody who is interested enough to read it!

I think the reason there’s a lot of confusion about the roles play in numbers is because there’s a massive conflation between the use of numbers in a media narrative sense, for KPI’s, and analytics.  I’ll explain it here now:

Media narratives:

This is the preeminent use of numbers in our footy society – to illustrate a point or further a narrative.  There’s nothing wrong with that.  If you want a shorthand way to describe how Caleb Daniel objectively had a career-best game against the Eagles, you can do so by saying he also had career-high disposal numbers.  If you want to demonstrate a narrative as to how the Hawks are good in a unique way, you can do so referencing contested possession numbers.

There’s nothing wrong with this in isolation.  In fact, there’s a lot that is good with this in isolation – as it increases the footy public’s interest in numbers.

The problem is when people conflate this to other areas of footy in numbers, because it’s not analytics.  Hawthorn don’t need to increase their CP wins, because as I explained last week, there’s a lot more that goes into understanding contested possessions.  The narrative isn’t the analytics here.

KPI’s

KPI’s aren’t analytics, because they’re not the statistical determination of how clubs win.

It’s also a shorthand way to measure effort and performance, but not the breakdown of how and why teams win.

If a club is below season averages in contested possessions at halfway through a game, it’s likely, in fact, scrap that, almost certainly because they are not working as hard around the contest than they have in other games.

For an analyst watching the game who “feels” like the club is not working hard working halfway through the game, the KPI is a good way to confirm that feeling.  This can then be communicated to the players at half-time to quickly demonstrate how they’re not working hard at the contest.

The problem with this is, again, in determining the sum of the parts that contribute to scores and wins, it’s not the same thing.  It’s context sensitive.

So what is analytics and why is it not numbers in the other sense?

Google tells me that sports analytics is the “management of structured historical data, the application of predictive analytic models that utilise that data, and the use of information systems to inform decision makers and enable them to help their organisation in gaining a competitive advantage on the field of play”

That seems pretty good, but remember, it’s not numbers to further a media narrative or to demonstrate KPI’s.

I’m surprised that people aren’t more distinct in the definition.  To simplify it – “using numbers to help teams win”.

In the AFL, and I’m surprised that more fans of numbers don’t understand this.

It is commonly accepted by footy analytics people that the starting point is:

  • The game first has to be divided into the sum of its parts, that is, that the game has a certain amount of possession chains.
  • These possession chains number 150-200 odd in number per team per game
  • These chains can be sourced from kick-ins, stoppages, and turnovers
  • League average wide, about 10% of chains end in a score
  • You can generate more possession chains in total by winning more than 50% of clearances

And then to add onto this

  • Improving your team’s likelihood of winning games and ultimately the premiership, is to:
    • Increase, collectively, your stoppage win percentage, and your points for per chain, and reduce your opposition’s points per stoppage chain.

This is then applied through a number of ways:

  • Measuring trade-offs, such as various relationships
    • The trade-off, and risk, in improving your defence/offence which naturally has the inverse impact on the other – if you attempt to move the ball more often through the corridor, for example, to make your offense better, you’re likely to make your defence worse by making your average turnovers in worse position. We can numerically analyse that data
    • The trade-off between stoppage and turnover, both offensively and defensively. If you aim, offensively to risk more of your chains ending in turnovers rather than stoppages or vice versa, it might make your net scores per chain better.  Likewise with defence – every club aims for a turnover, but you have to accept the trade-offs (like attempting to intercept mark every time for a turnover rather than attempting to spoil – you have to coach your players to make optimal decisions)
  • How wider tactical systems, player recruitment, player selection, development methods etc. ultimately impact these efficiencies per chain and therefore whether you win or not.

Ultimately, however, the purpose numbers in an analytics sense is to use numbers in relation to chain efficiency.  When looking at Caleb Daniel’s disposal numbers, one has to look at it in relation to how he improved his team in chain efficiency offensively and defensively, which can be determined with various numbers.  That’s the thing that all good analysts are relating numbers to – how a player ultimately helps or hinders his team in the efficiencies that are the sum of the parts involved in winning.

There’s plenty of more things involved in analytics – player projection systems into the future, using player tracking data, understanding how systems management works in an analytical sense, what is the proper determination of a replacement players, what is the proper baseline for a “marginal” contribution etc. etc, but I said this would be a short article!

So it’s clear to see how numbers from an analytics sense differs from numbers in a narrative or KPI sense.

The problem is there are people who try to have a foot in each camp – like the figurehead of the footy media industry’s numbers movement in David King.  He’s made suggestions from a media narrative sense that the Hawks are good with low CP numbers, nothing wrong with that.  But once he branches into analytics and talks about how their CP numbers impact their ability to win, he has to look at how it impacts things like clearances, score efficiency from clearances and so forth – the fundamental basis for analytics in footy.  The issue becomes incorrect and confusion due to this media narrative/analytics conflation, and as a result, the public is actually confused and has a greater distrust in footy stats which feeds back into the clubs.

Also, lets not not recognise the role that Champion Data play here.  As they’re the guardian of numbers, they often push forward how the public views numbers.  But they also have a foot in each camp – as their revenue sources is both the media and the clubs and they spend a lot of their time servicing the numbers they send to the media.  The very creation of Champion Data is because of not so much a club ‘moneyball’ need of stats, but rather a fan and media narrative for stats – which means that how we view numbers in a narrative sense exists to this day.  Not that there’s anything wrong with Champion Data doing so, and I’m not criticising them for servicing the media, it’s just something that needs to be understood – it’s up to the AFL to decide whether to publicly release stats of for clubs to do more ‘moneyball-esqe’ analytics in terms of tactics and list management, not Champion Data to do it for the club.

Don’t me afraid to send to tweet me/send me a private message on Twitter if you’re interested in hearing more!  I’m happy to talk!

Geelong vs Hawthorn tactical preview

Friday night, wet footy in front of a mass crowd MCG.  Should be an absolute cracker.

Hawthorn and Geelong last matched up in Round 1 this year.  Geelong won that largely on the back of Patrick Dangerfield who ended up playing one of the greatest years by a midfielder of all time.

Hawks pressure at stoppages vs Dangerwood out of stoppages

Whilst Selwood helps, Dangerfield redefines the boundaries of what players can do out of stoppages.  His ability to win contested ball and gain metres out of stoppages is absolutely-bloody-phenomenal.  Out of this world.  This can be seen a lot out of centre bounces, but tactically I’m not going to talk about it because they’re largely random and the Hawks recognise that.  Instead I’m going to talk about it from the perspective of stoppages around the ground.

I really can’t emphasise this enough.  Everybody knows how good Dangerfield has been, but for those who don’t watch many Geelong games or don’t critically analyse the game at all times, the point really needs to be made – Dangerfield’s raw power in congestion is the fulcrum for much of Geelong’s ball movement and ability to drive forward.  His contested-and-metres gained combination is unparalleled in the league.  Of course, this isn’t mentioning the core strength through the hips, the ability to attack the ball at speed and his tendency to play for free kicks by exaggerating contact.

Dangerfield does this by having great awareness and poise around the stoppages.  Look at how he slips free of Robinson who is meant to be tagging him here:

pu8i2zc

Geelong out of stoppages also structure in such a way to allow extractors such as Joel Selwood and Dangerfield to get first hands on the ball, players in particular such as Mark Bilcavs using his size blocks space and uses his body when he’s not going third man up to make it more difficult for their opposition to win the ball, but not just limited to him, as plenty of other Cats are in this mindset:

qld7ys2

Notice how the two Geelong players on the right hand side of the ball in this image are holding back the Brisbane players and blocking them to allow Dangerfield to run in that direction.  They certainly do what they can to work in Dangerfield’s favour.

How does Hawthorn match up with this?

As I discussed in my piece about contested possessions, Hawthorn are dead last for pre-clearance contested possession, but ranked first for pre-clearance pressure around the ball.

They do this by having a defensive mindset around stoppages, setting up to lose the stoppage in the first place by placing players on the defensive side so they don’t get caught outside.  They don’t actually aim to win the ball, rather than pressure it on the outside – here is a typical example:

hgy4oyn

This is immediately after a stoppage get by the Pies.  It’s not hard to see the encirclement at hand here.

The interesting thing to see is that the Hawks aren’t necessarily the best at defending chains from stoppages once the opposition get through with the clearance in a points per clearance point of view, it’s just that they’re extremely good at stopping opposition first hands from breaking open the game completely.

This is going to be really, really interesting at the stoppages in this final.  The Geelong strength of powering out of stoppages largely through the power of Dangerfield and even recently Sam Menegola vs the nature of Hawks to pressure the first hands of the opposition.  Will Dangerfield continue to power through and clear the ball, or will Hawks pressure be all too much?

And of course, Dangerfield can influence the game other ways, like resting forward and being an overhead marking threat.  It’s reasons like that why the Cats are favourites for this game!
Geelong defensive structure vs Hawks ball movement

When Geelong kept Adelaide to absolutely minute scores from turnovers in their two matchups this year, well below the Crows’ season averages, it was no accident.  The Cats have an extremely tall, yet compact defence full of experienced players who all know how to play off each other and have developed extremely good defensive chemistry.

I apologise for not having images for this because with TV broadcast cameras it’s hard to illustrate the point.  When Geelong turn the ball over from defence or through midfield, the innate chemistry of the Geelong defenders and their size and ability overhead, and the seamless introduction of players like Jake Kolodjashnij in the defensive unit have meant that teams cannot counterattack with the ruthless efficiency against the Cats like they do other teams.

It’s not just against their own turnovers when this chemistry helps – it’s also against the small, incremental ball movement – Hawks strength – where they’re also strong themselves.

This cohesion in defensive structure makes it harder to slice through the opposition and find uncontested marks, forcing long kicks to the taller defence which plays in the Cats’ hands.  This is a perfect example, where the taller Cats are going to impact the marking contest:

ane2f8n

However, the slowish nature of the Cats defence means that if their forward pressure if off, or if they get exposed with rapid ball movement through run off half back or quick switches across the ground, which Hawks are certainly capable of.

This matches up well with Hawks incremental ball movement which is among the best in the league:

xvssyvc

This is something that’s typical of the Hawks – moving the ball wide to spread the opposition zone defence from a turnover to try and find holes.  This constant movement, like a motion offence in basketball, forces opposition quick thinking and allows for gaps where passes can be made.

I think that the Hawks ball movement could expose some of the slow nature of the Geelong structure and spread them wide on the expanses of the G, however, the wet game plays straight into the Cats hands as the skill level drops and the raw power of their inside midfielder come to the fore.  However, I also think that the Cats might lack the ground-ball players forward and back outside of their midfield unit and clever players from the Hawks who are good at clearing the ball in congestion could prove the difference.

Certainly going to be a great game!

The Contested Possessions Myth

The common perception of “win contested possessions, you’ll win the game” is something that is found everywhere in the AFL landscape and its actual importance is in reality much, much lower.

I consider myself a football sceptic.  It’s what I try to blog about from an analysis perspective, whether it be looking at the game from a tactical or statistical analytics point of view.

Google defines sceptical as “not easily convinced; having doubts or reservations” which is usually against the commonly accepted viewpoint of something.

In football terms, the commonly accepted viewpoint of football as a whole is driven by the media which is dominated by a white male viewpoint, driven by the “boys club” of ex-players in terms of analysis (seriously, just listen to Cameron Ling’s ‘special comments’) more and the crusty private school background in terms of administration.

The problem with this from an analysis perspective is that you get an echo chamber of analysis through the application of statistical analytics.  The viewpoint is often being driven by somebody with no expertise in an area whatsoever trying to divide the sum of the parts of a game through misapplied numbers.

This brings me to David King.  I’m sure King is a lovely man who is extremely passionate in being ethical in his role in football media, but that doesn’t mean he’s an expert on the relationship between game statistics.

The fact that King is seen as some sort of expert in the statistical analysis of the game absolutely staggers me.  I genuinely do appreciate that he’s attempting to be more forthright in using numbers and video footage to analyse the game, and that is certainly something that we need more of in general.  But come on! As far as I’m concerned he’s a bloke who played a lot of footy in the 90’s era, grew up in that anti-intellectual culture, was a failed assistant coach and has now made a niche for himself in the media as being some “stats expert” despite having no genuine expertise in statistics.

This brings me to the topic of here today – contested possession.

Among a game that money is being pumped into to try and scientifically solve to win, there’s probably no single statistic that’s being quoted more and more as the barometer for success than contested possession.  I think it’s a myth and the footy “analysis” media has us all deluded.

Let’s first establish how much the concept of winning contested possession permeates through footy analysis across the landscape.  Take a typical episode of Fox Footy’s On the Couch, a show that – Ryan Buckland at The Roar and Andrew Faulkner at The Australian both opened up articles on contested possessions by quoting members of the panel as saying “win contested possession and you’ll win the game”.  I’m not too worried about this, because it’s from dumb ex-players and means that the public is subsequently dumbed down and it makes analytical thinkers like myself seem more elevated, when in reality I’m just some young kid who only distinct feature in regards to footy is that I’m questioning, not accepting, what people say about the game.

The problem is that when you have people who should be more forthright about the value of contested possessions – people who work with footy stats every day – reinforcing this myth, I scratch my head a little bit.

Glenn Luff, who as Champion Data’s head honcho literally works as a full time job to give meaning to numbers in a football contest, appearing on radio and writing the AFL Prospectus, has bought into this myth.  Jacob Wilson, the Champion Data employee who was featured in the recent Vice Sports/Draftstars feature on footy stats, also talked about how important contested possessions are to winning.

Even the godfather of stats, Ted Hopkins, isn’t immune!  From the Andrew Faulkner article:

“Winning contested possession is critical to winning games”

Sorry, Ted, I love what you’ve done for the game but that one really hurt.  Although he has also said contested possessions are overrated, so I don’t know what to think!

 

Karl Jackson who is the main technical guy at Champion Data also reinforces his beleif, which is also a good segue to talk about the technical mathsy impact of contested possessions this in response to Ryan Buckland’s piece:

In response to Cameron’s comment above (and indirectly to his article) margin does increase with a bigger win in contested possessions but not perfectly. Contested possession differential and final margin have a correlation of about 0.63 this year, or an R-squared of just under 40%, meaning that 40% of the variability in final margins across single matches can be explained by the contested possession differential alone – not bad for a single stat. Obviously other things play a part in the result of games – ball use, pressure, accuracy in front of goal, etc. but contested possessions are a good starting point.

As for the value of a contested possession, theoretically to start with, roughly 10% of a team’s possession chains lead to a goal. Each contested possession is an opportunity to either maintain a possession chain or start a new one by turning over the ball, so with one contested possession you’ve given yourself an opportunity to score (10% of 6 points = 0.6 points net) and have prevented the opposition an opportunity to score (10% of 6 points = 0.6 points net) so one contested possession is worth 1.2 points. Obviously these scoring rates change based on where the contested possession happens, but hopefully you follow the logic.

As for practically – a simple, crude linear regression of margin against contested possession differential puts a value of 1.6 points per contested possession for this season. So no, one contested possession across the course of the game is rarely going to give you a significant advantage, but win/lose the stat consistently or by big margins and you’re in trouble.

 

Moving on, I haven’t really explained why it’s a myth.

Okay, so I admit, I was flapping my gums about it being a myth, and I need to back up these against-the-grain views with some proper rational and considered thinking about contested possessions.

Here it is!

Lets define contested possessions

There’s a bit of inconsistency over what the technical definition of a contested possession is online, and it generally seems to fall into these two similar categorises:

  • Winning the ball when members from both teams had the opportunity to do so;
  • Winning a ball in dispute

I’m not going to get hung up on what the definition of a contested possession is and whether the definition and therefore match-day calling of it by Champion Data needs to be loosened or tightened.  I don’t understand the processes in statistical definitions or how Champion Data calls matches, so I’m going to move on.

On average in a home and away game in 2016, there were 284 contested possessions (so an average of 142 won per team) which compares to 466 uncontested possessions. (All statistics sourced from AFL Tables unless otherwise stated).

Every single team in the league for the 2016 home and away season won between 52.5% and 47.0% of the total contested possessions within the games that they played, with the Bulldogs and Hawks winning those amounts respectively.

This is important to remember because I will be referring back to this later.

Moving on.

Correlation does not equal causation.

Correlation does not equal causation.

Correlation does not equal causation.

The reason that many people reinforce the myth that winning contested possessions contributes to wins is because it’s very strongly correlated with the final margin of games.  Further insight from both Ryan Buckland and Karl Jackson’s response explains this in further detail.

So jump to conclusions and believe that if you should aim to win contested possessions to improve your chance at winning.  But that’s a myth because correlation does not equal causation!

Going back to Karl’s response earlier, a team would think that winning more contested possessions such would lead to a higher margin, and teams should win more contested possessions because that leads to more possession chains and more goals.

But does it really?

I can’t disagree with the concept that contested possessions and margins correlate strongly.  But why do they correlate strongly?

Because I think teams that are good at winning contested possessions are just generally good football teams who do the other aspects of footy well.

Teams that win contested possessions are made up of players or tactical systems that are good at winning contested possessions.  These players and these tactical systems also are good at ball movement or simply getting into the head of the opposition.  That doesn’t mean that the contested possessions in and of themselves are contributing to a higher margin.

Why?  Because I believe teams that are good at contested possessions are doing other things well.

Teams that are good at contested possessions are usually:

  • Filled with players who simply are good players, so players who win at contested possession are also kicking more accurately, understanding structures and just generally good at football in terms of leadership, experience, fitness, physicality and so forth.  This the team being the sum of the parts of its individual players
  • An increased margin in contested possessions usually means that a team structure as a whole is also good at other aspects of footy. Because contested possessions means that you’re a team that are winning more balls in dispute (as per the definition above):
    • That probably means your team is working harder and running harder, or fitter with sports science, or doing good rotations both through the bench and player positions. If you’re winning more contests, it probably means you are the fitter team, or the team that gets good mis-matches tactically, which means that you’re also going to be better at other aspects of the game given you’re the fitter team and the structures balance to you well.

If that’s hard to sort of get your head around, think of it from other sports, like NBA Basketball.  Nylon Calculus introduced Dredge, a play-by-play box score metric for Basketball, it found that players who commit technical fouls tend to be better defenders and therefore a positive value is given to committing technical fouls in this player valuation system.  On face value, that’s ludicrous – the nature of committing a technical foul shouldn’t be a positive thing, because committing a technical foul gives your opponent free throws and possession and therefore in a vacuum is a bad thing and should be negatively weighed.  But players who commit technical fouls tend to be active on defence and better defenders, and this outweighs the relatively minor material impact of these defenders committing more technical fouls.  Now think of that in terms of contested possession – on a team level, it isn’t the contested ball winning nature itself that is contributing to the final margin, however, it’s how it’s a team that is winning contested possession at a greater than 50% rate is the representation of a team that might any one or more of: being more physically stronger in general, running harder, getting its tactical opponents matchups right or filled with 22 players who are good at all aspects of football above and beyond simply winning contested ball.  That’s why the correlation is so strong.

This also results as the fact that coaches use contested possessions as a KPI – but KPI’s are not analytics!  KPI’s are just a way to communicate to the team that they need to work harder in the contest.  If a team is below their season averages in contested possessions, the coaches are probably disappointed because it’s a short-hand, proxy way of realising that the team is playing poorly in the contest.  But analytics – figuring out the process towards scoring, and preventing your opposition from scoring – is not KPI’s!  Even though they both use numbers, they’re different things!

So now we’ve discussed how correlation does not equal causation, now let’s look at the direct material impact of winning the contested ball.

Like discussed before, 2016 had 284 contested possessions per game.  2016 also had 73 clearances per game across both teams.  We also know that 96% of clearances have a contested possession before immediately beforehand.

If you’re with me on the maths front, 96% of 73 clearances per game means that 70 of the 284 contested possessions per game is directly related to the clearances at hand.  That’s a significant figure – one whole quarter – in that contested possessions are in a pre-clearance mode and therefore don’t matter until there is a clearance, but wait, there’s more.  That assumes that in pre-clearance mode, there’s an average of 0.96 contested possessions per clearance when in reality the number is much, much higher.  This is because there’s repeat stoppages a not insignificant amount of the time, and there can be multiple contested possessions before a clearance at a stoppage, given dispossession, pressure, tackling, pilfering and so forth.  So it’s higher than 25-odd%, but probably lower than 50% but we don’t really know.  This by the way, is why we need to #freethestats, so that critical differences in contested possessions, pre-clearance and post-clearance can be made public and this changes the football public’s understanding of key topics like contested possessions.

The concept of pre-clearance and post-clearance stats or contested possessions is critical for football analysis.  Absolutely critical, because of the very nature of dividing the game up into possession chains, many of which are sourced from stoppages.  The concept of pre-clearance and post-clearance stats isn’t foreign to Champion Data.  It’s featured in the AFL’s live text updates in their online match feeds, on Fox Footy match previews and Champion Data certainly knew what the hell I was on about when I asked them.

In summary, to analyse the impact on contested possessions, you have to split the division into pre- and post-clearance, with the pre-clearance contested possessions only part of the story.

Pre-clearance contested possessions, which make up more than one quarter of all contested possessions, don’t matter unless we can analyse an impact on how it impacts clearance win differential and net equity difference from clearances, which can be roughly measured by scores from clearances.  This point is overlooked by virtually everybody who comments on contested possessions in a public sphere, until now.

It’s common sense to suggest that when we look at how good or bad a team is at contested possessions pre-clearance, we shouldn’t look at them in raw numbers.  Rather, we should look at how it impacts clearance win percentages, and the differences in scoring from stoppages or the efficiencies in possession chains sourced from clearance wins.  Given that, despite their 18th ranking in pre-clearance contested possessions, Hawthorn aren’t ranked anywhere near that lowly for those two metrics, they’re clearly not to worried about increasing their ranking from number 18 in pre-clearance contested possession, that makes up perhaps a third of all contested possessions.

Contested possessions in pre-clearance mode work hand in hand with pressure.  Hawthorn are ranked dead last in pre-clearance contested possessions.  This means that they are the worst team at getting “first hands” on the ball after the initial hitout.  But the thing is, it doesn’t matter – Hawks pressure better than any other team in the competition in pre-clearance mode (and just as an aside this is why Liam Shiels is underrated, as he’s the most critical player for the Hawks in this regard).  This then sort of evens each other out, because the fact that their opposition get first hands on the ball more often against the Hawks is limited by the fact that those first hands are more pressured than against any other team.

Hawks are among the league’s leaders at repeat stoppages, so logically you can dictate that they force repeat stoppages when the opposition gets first hands at the stoppage better than any other club – over 10% of the hitouts in their games results in no clearance.  This “first hands” that the opposition wins, which contributes probably somewhere near enough to half of the Hawk’s poor nature in winning contested possessions, contrary to popular belief, is minimised to a minute extent if the Hawks are forcing repeat stoppages at a significantly high rate above league average.  Then to expand on this, their actual clearance differential ranking is higher than 18th in the competition like their pre-clearance contested possession ranking, which means that it doesn’t matter.

Then chuck in the fact that the Hawks are only a minor net negative in differentials in scoring from clearances (per clearance win) – despite not investing any effort into getting first hands on the ball shows you how much of a myth winning contested possessions are.  The Hawks can already invest more time and resources in terms of tactical analysis and training away from this pre-stoppage scenario and more toward ball movement and kicking, which they’re among the league’s elite.

I don’t really want to guess as to guess as to what clubs are doing from an analytical sense, but Darren O’Shaughnessey works as an analytics consultant at Hawthorn.  Given that he’s already publicly shared the random, don’t-invest-resources nature of centre bounces that he’s worked with at the Hawks, it wouldn’t surprise me if he’s come to the same conclusion as me above and helped direct the Hawks into investing less resources about worrying in winning contested possessions in a pre-clearance setting to invest resources elsewhere.

I don’t mean this to mean that Hawks, in a vacuum, would not prefer to be better out of clearances or win more first-hands on the ball at stoppages.

Because you’re more likely to get a clearance if you get first hands.  Winning clearances is good because it gets your more possession chains.  I’m sure if there was no trade-off, the Hawks would aim to improve their pre-clearance contested possession numbers.  I’m sure the Hawks don’t want to be ranked 18th in contested possession win rate, because in a vacuum, winning more contested possessions does indeed contribute to more wins.  But there is a trade-off.  This trade-off would involve multiple things, such as the possibility of the Hawks dropping further from rank 1 in pre-clearance pressure than it would rise from number 18 in pre-clearance contested possession (therefore having a negative net impact) or being worse at defending clearance losses (say by swarming the stoppage more and pushing defenders up around the stoppages), and so forth.

So no, David King, Hawthorn don’t need to improve their contested possessions in order to win the flag.  It’s a lot more complex than simply their differentials.

So what about post-clearance contested possessions?

For entirely different reasons I also think that it doesn’t really materially matter where a team is ranked in contested-possessions post-clearance.

I don’t mean that in a literal sense, because obviously if you win greater than 50% of balls in dispute in general play you’ll generate more possession in general and an ability to score.  Again, in isolation and in a vacuum, winning more contested ball is a good thing and will help you win more games.  I just think that the differences between teams are so minute, and the impact minimal, that too much time and thought is given to this than people realise.

I think that every club is simply so close to 50% anyway that the difference in being ranked, say 5th and 15th doesn’t really matter in winning a game compared to much more important things such as tactical structures, accurate kicking for goal and so forth.  Yet, the effort into improving your ranking gets over-emphasised so much by the media for the sake of a very minor amount of material contested possessions per game.

Contested Marks are a contested possession and whether you win them largely depends on whether you pick a lot of tall blokes in your team, whether you kick to packs or good old-fashioned randomness.  So dilute contested possessions by, say, a third to remove the pre-clearance ones, and then get rid of the 23 contested marks per game.  You’re left with 167 contested possessions for both teams per game that are post-clearance, and are not contested marks, if you make the very liberal assumption that 2/3 of all contested possessions are post-clearance.

If you won 51% of the contested ball, you’d be ranked 5th in the competition.  If you won 49%, you’d be ranked 14th.  That’s quite a large difference in ranking, but it’s only a minor difference in percentage.

I understand that these percentages are for all contested possessions, not post-clearance ones, but as the stat’s aren’t publicly avaliable, we’ve got to make the assumption that the percentages hold true for post-clearance and there isn’t a significantly different range for pre-and post-clearance contested possessions.

What’s 49% of 167? 81.8.  What’s 51% of 167? 85.2.  That’s a difference 3.4 contested possessions per game.  What’s material difference of 3.4 contested possessions per game can make? Not much!

Is it really worthwhile working so hard from an investment of footy department resources to improve from 14th to 5th on the post-clearance contested possession ranking to gain a very small handful of more contested possession wins?  That sort of investment – through more training investment or video analysis training of your players – has a trade-off in terms of investing more time in letting them video analysis tactical structures and so forth.  The material impact of being ranked highly or lowly in contested possessions doesn’t really matter, because all teams are so close to 50% anyway.

The more high-various impact on the final results of games is Shot Quality Production, which is a end result of attacking and defending ball movement.  Ball movement.  Ball movement. One more time – ball movement.  Just quietly, 90% of the game is ball movement.  Every type of analysis and coaching to win games should be done in relation to ball movement.  But just quietly, because many people in footy haven’t caught up to this principle.  The Shot Quality Production/Expected Score is something that people who know what they’re talking about – like Darren O’Shaughnessey – use, rather than people like David King who will bang on and on about “winning contested possessions”.

So there you have it – the myth of contested possessions.

There’s plenty more that I could have talked about, but this is already over 3500 words!

I hope this was worth and you gained some new perspective on the game we all love!

Twitter: @MadnessOfSport

Women’s All-Star game tactical review

Over six thousand people attended the Women’s All-Star game at the Whitten Oval and whilst the fans saw a high quality game that has done wonderful things for the development of the women’s game, the battle on field from a tactical perspective was also very invigorating for the fans.  The sixteen a side nature of the game and smaller ball successfully made for a high-scoring affair that in many ways was a symbolic end to the exhibition era over the last four years, with a transition to the national league next year.

So how exactly did the Dogs turn around their fortunes from their previous all-star battles and win this game?  They did so through a more balanced line-up and building upon their strengths as a team more than Melbourne did.

First consider the impact that removing four players from the field would have had from a tactical perspective, that may have played into the Bulldogs hands.  Although the players lined up nominally without wings, each of the teams played with five in the forward line.  This may have meant that the taller forward line of Harris and Frederick-Traub of Melbourne, who excelled more in contested marking rather than moving in space, gained less benefit than a player who excelled in leading into space in Hope for the Dogs up forward.

How were the Demons strengths diminished?

The strengths, at least, on paper for the Demons were:

  • A better collection of midfielders, both inside and out, led by Daisy Pearce, but also including marquees Emma Swanson, Kellie Gibson and Kara Donnellan
  • Better tall forwards, let by Tayla Harris and Sabrina Frederick-Traub

In summary though, they had Daisy Pearce who is head and shoulders clearly the best female footballer in the country.

So, firstly, the first point of order was to stop Pearce winning masses of the ball on the outside.

Pearce has dominated at all levels of footy that she has played in recent years, topping out with over 30 possessions and over 200 Champion Data Rankings Points (something that no male player has done at AFL level) in multiple exhibition games.  In those games, it was her ability to combine inside and outside work – for example she took eleven marks in the exhibition game against Brisbane – which made her such a destructive player.

Pearce was much quieter in this final exhibition match.  Although she won plenty of the ball, finishing up with 25 touches to be only one of two Demons players to have more than 18, her influence on the game was muted – only two clearances and one inside 50 for the entire game.  It’s doubtful whether that was simply Pearce having a poor game or whether it was a tactical scheme by the Dogs, but they did make an effort to pressure her.  An example of that is here:

do8pfm9

Here Pearce wins the ball in congestion but is forced to move the ball quickly given the pressure from Katie Brennan.  She forces a handball which becomes a turnover.

Another factor was the goal-kicking of the tall forwards Harris and Frederick-Traub.  Early, these two players looked dangerous, taking more marks inside 50 and looking like a key position duo that could cause some harm, however, both played missed some easy set shots.

The other issue with these tall forwards is that if they weren’t kicking the ball from marking it, they were not providing much forward pressure in contrast to the smaller forward line of the Dogs.  For example, Harris laid zero tackles for the entire match.  Brianna Davey was slightly undersized but she impressed winning the ball at ground level and rebounding, in particular in the second and third quarters, and the Dogs captain Steph Chiocci was able to rack up numbers running out of half back with a bigger forward line.

km9vpzi

This is a typical example off the type of the lack of forward pressure offered by the Demons forward line.  Harris is the player closest to the ball, yet there are four Dogs players around her looking to run past her – including those zoning off the other taller, yet slightly less mobile player in Frederick-Traub.  This lack of forward pressure cost the Dees plenty here in this game.

Another example of Chiocci winning the ball in defence and using it well, handballing it off to Davey:

bmiflde

The Dogs also zoned off across half-back and forced rushed kicking that often led to intercept marks or possessions rather than the Dees being able to pin-point kicks to these tall forwards.  Here, Frederick-Traub is unable to influence the contest and the ball is intercept marked by the Dogs half-backs zoning off from the other side of the ground:

wdewrjx

This is an example of generally poor ball use moving the ball inside 50 by the Demons, who actually led the inside 50 count on the match but turned it over or didn’t make the most of these entries into an open forward line.

The Demons defensive unit also got sucked into pushing up the ground too highly in order move the ball forward, and thus were poor at covering the space in front of Mo Hope at full forward:

What were the Dogs strengths?

While most people had players like Ellie Blackburn and Katie Brennan as the Dogs’ best, and they played excellent games, they weren’t as directly influential to the structural successes of the Dogs.  That was largely due to Brianna Davey at full-back, Emma King in the ruck and Kaitlyn Ashmore as an outside midfielder.

Blackburn was excellent inside and used great endurance to link up all over the ground.  She was a large part in the large clearance differential in the match, Dogs winning that 35-26, however she wasn’t humungous part of the ball movement heading forward and in fact at times was poor by foot.

Brennan was clearly a superstar for the Dogs.  Working hard through midfield, as a deep forward and as a high half-forward leading up the ground, she demonstrated great physical attributes from an endurance and strength perspective.  She laid eight tackles, won plenty of the ball and kicked a couple of goals in what was probably the best performance.  However, this isn’t a point of difference focus.

The first point of difference was Brianna Davey out of defence.

The Carlton marquee was instrumental for the Dogs with rebounding and clean use out of defence.  Here is an example of her attacking the ball and rebounding it for the Dogs:

sunxnhc

Davey is an interesting prospect because she has many traits that are atypical of a key defender – she demonstrated very good ball use under direct pressure.  She has a bit of a lack of height and perhaps isn’t brilliant being directly accountable, but her ability to read the flight of the ball and rebound and intercept to get the ball moving was brilliant.

The second point of difference was Emma King in the ruck.

King is the best ruck in Australia and it was clear to see why here.  Dominating around the stoppages (and contributing to massive hitout and clearance differentials) she was a key factor around the ball and structuring up around the play as well.  Her running ability was excellent for a ruck and she was involved in link up play through ball movement like that can be seen here, when Ashmore kicks it to a leading King (just inside the centre square in the picture):

ig4ekgq

One other final point of difference is that this example also demonstrates is that of Ashmore, her being quite clearly the most explosive outside player who took full advantage of the wider expanses of the Whitten Oval.  She used her pace and athleticism to great effect (even if her kicking was a bit off at times), breaking into space and providing a great switch open, leading the Dogs in inside 50’s for the match.  Whilst she only won the ball sixteen times, it was her contribution to the ball movement for the Dogs that was the clear difference.

The game had some very astute tactical coaching from Paul Groves (the first graduate from the Luke Beveridge school to coach an AFL team anywhere, male or female, playing under him at St. Bede’s Mentone), such as his use of Ashmore on the outside and playing Brennan more through the middle (and leaving space for Hope up forward), which will be very interesting to see if he continues to demonstrate that in the inaugural NWL next year.