Followers of this blog have likely read my thoughts on the lacklustre state of AFL analytics, both inside clubs and throughout media has a whole. The obvious question that comes to mind after reading those pieces is: If analytics and stats aren’t being used to their full ability right now, what exactly would *you* do to use them any better? This series of articles hopefully begins to provide an answer to that question.
These articles are framed as an “analytical preview of teams before each final”, but in reality I will be using them as a way to introduce some new concepts, visualisations and ideas about the general direction in which I think good footy discussion should be moving. (If you’re looking for in-depth, tactical discussion ahead of each match, Nick Welch at Madness of Sport has been doing some good work.)
Each article will have something you haven’t seen before, patched together from rough work I have been doing over the last few months. Hopefully you can see the wide-reaching applications for some of this stuff as well as learning a thing or two about the AFL finalists
Other posts in this series:
Adelaide v North Melbourne (Shot Locations Heat-Map and Age Profile)
West Coast v Western Bulldogs (Mark/Free Kick Mapping and Player Tracking)
Sydney v GWS (Forward Reliance/Leading Areas and Team Ratings)
Geelong (Mapping Score Involvements)
One of the few “secret sauce” statistics the AFL makes publicly available through their app is the number of “Score Involvements” for each player. A score involvement, as far as I can tell, is simply registered as player having touched the ball and contributed to the possession chain at some stage in the lead-up to a score. Here’s how the 2016 season totals look:
That list is unsurprisingly dominated by forwards1. This makes sense, your own shots at goal count towards score involvements. If you are taking 100+ shots a year, you’re going to have a lot of score involvements.
Once we strip out the specialist forwards, we’re left with only two guys. The highest ranked of these being the almost unbackable Brownlow favourite, Patrick Dangerfield.
I wanted to get an idea of where on the field Danger has done most of his work so I developed the following visualisation:
Each one of these lines represents a score involvement, starting from where he disposed of the ball and finishing where the next player took possession2. You can think of these lines as the paths of any kicks/handballs/knock-ons Dangerfield made which eventually lead to a goal. The colours represent whether the disposal resulted in a score for himself, a score for the player to receive the ball, or a score for the opposition from a turnover.
Now, I admit that’s a lot of lines. But let’s go through this and see what we can learn:
- Most of his score involvements come in the forward half. This isn’t really a shock for an attacking midfielder like Danger, but only 7 times this season has he touched ball in defensive 50 and scoring shot come from the resulting play.
- He can make the length of the MCG in about 3 long kicks.
- Turnovers can cost you from anywhere on the ground. For all the touches he has had, Dangerfield has only had points scored against him from his own turnovers 9 times this season. That being said, lots of these were lost in his own forward half which highlights the big importance turnovers have in the modern game.
- Most of his involvements are either direct assists or shots. 99 score times this season points were scored either off Danger’s boot or by the player who got it from him.
- Most of his assists are from outside 50. Dangerfield has the most inside 50s in the comp by a country mile. We can see here that a good chunk of these get converted into goals. To make this a bit clearer, I’ll highlight all the assists:
- More often than not when inside 50, he’ll go for goal. By far, the majority of Danger’s SI from inside the arc are him scoring himself. This is not necessarily a bad thing, but certainly something to be wary of when playing on him. You can see from his ShotPlot that he doesn’t mind the long range effort.
Whether any of this provides us any insight into how Hawthorn can stop Dangerfield on Friday night is up for debate. But you’d have to think that if he gets around his average of 8.7 score involvements a game, Geelong will be hard to beat.
Hawthorn (Close games and SQP)
There was a very interesting tweet yesterday that caught my attention:
If the AFL ladder was based on Champion Data's expected score. Few interesting changes: pic.twitter.com/c41GQsy2hb
— Ethan (@ethan_meldrum) September 7, 2016
This is how the ladder would look if every result had gone by who created the highest quality chances in the match. Champion Data call this the “Expected Score”, but the concept is identical to my “Shot Quality Production”.3
What we can see is that Hawthorn won an extra 4 games in which they had not in fact created the better chances. These wins have been due to a combination of straight-kicking on their part (which is inherently luck based), and poor kicking kicking by their opponents. We could make the argument that this is what makes Hawthorn a “great team”, but it would be a stretch to say they are so great that they can put off an opponent kicking for goal from the same spot on the pitch better than any other side. It’s very likely that luck got them at least some of these extra wins.
Hawthorn have also won the 6 games this year that they were within a kick of losing just seconds before the siren. The narrative seems to have emerged that Hawthorn just find a way to win the tight ones. But it seems more likely to me that a huge dose of luck went into Sam Mitchell swinging his boot out of the ruck to go 50m downfield into Gunston’s hands against Sydney in RD17, or even Jack Fitzpatrick getting space and a kind bounce to kick a goal from the centre-square last week again Collingwood.
The research into this backs me up.
I’m not saying Hawthorn are not a good side, I’m just saying that on the balance of probability, they are lucky not to be over in Perth playing an elimination final tonight.
Good luck hearing anybody make that argument amongst the cries of “Fourthorn”!
- Even more specifically, forwards named Tom Lynch
- Or in the case of shots, finishing at the goal-line
- My system also used to go by the name “ExpScore”, but I’ve since changed that in order to make clearer that what we are actually judging is shot quality, and so that we no longer need to rely on the strict mathematical definition of “expectation”.