- Video
-
Club Roar
Captured a great grassroots sporting moment? We want to see it!Content Collections
The Roar Community
- Join
- Login
Ever since reading Moneyball in the early 2000s, I have been preoccupied to varying degrees to try and quantify the various elements of AFL football.
Aside from finding it interesting on a personal level I do think there is scope to understand in a meaningful and measurable way the following things.
I think these things have application for list managers and teams but at a simpler level, provide a way to talk among fans that help provide insight and complement the way we interpret the game when we watch it.
Ultimately this can shed light around accepted tropes like ‘big men take time to develop’ or getting a sense for what fairness looks like in a trade or free agency signing.
Recently a question that has come up for me again is to try and evaluate the scoreboard impact that specific players and more broadly and specific type of player can have. For example is a Tom Mitchell 53-touch performance more valuable than Lance Franklin having 20 touches and six goals?
What about a typical Cyril Rioli performance? As in, a low possession half forward who tends to create scoring opportunities when under intense pressure?
I don’t suppose to have the answers per se – however I do like to work to try and come closer to whatever the answer is. In US sports, basketball and baseball in particular, analytics are more prevalent than ever before. One of the pioneers in NBA Analytics is Dean Oliver.
Amongst other things, Oliver developed metrics for measuring offensive and defensive efficiency that derive from the typical box score but give more nuance. They are known as four factor offence (and four factor defence) I have tried to use this approach to look at both offence and defence in AFL teams and individually for players.
I should say that in undertaking this type of investigation I appreciate that some results might challenge your subjective reading of the game and what constitutes a quality performance, and that for a range of reasons these measures are imperfect. Nevertheless let me try and explain four factor offence for AFL teams and players.
Effective scoring rate: In relation to basketball, effective field goal rate weights two and three point scoring against total attempts to properly evaluate the difference between types of scoring.
For AFL I have used a similar approach by giving six points for each goal, one point for a behind and dividing that by the number of kicks the player had. Therefore a player with three goals two behinds and ten kicks would have an effective scoring rate of two (3 x 6 + 2 / 10 = 20/10 = 2).
The thing I don’t like about this stat is that in basketball, even defensive players are still likely to put up at least an attempt and as such the measure applies to everyone. Obviously we don’t expect certain players to contribute in this way so there are some obvious inaccurate results but, from an efficiency perspective, I think this makes sense for offensive or midfield players.
Turnover rate: Fairly simple – number of turnovers divided by total disposals.
Advancement rate: This doesn’t have a corollary with basketball’s four factors but as a statistic it correlates very strongly with points scored. Ultimately I’m taking the metres gained by an individual and dividing it by the total metres gained by the team.
Offensive retention: I thought of this as comp for rebounding in basketball. If a rebound is taking possession where the ball is in dispute then I thought I would try and look at aspects of performance that either generate scoring or retain possession in offensive situations. There is some mathematical manipulation I use on this metric that I won’t core you with but in essence it compiles goal assists, tackles inside 50, inside 50 entries and marks inside 50 as a proportion of the number of forward 50 entries the team has.
Together I put these four factors into a multivariate linear model and produced a result that correlates strongly with team scoring with very little error. In short without explaining the different weightings the equation is; Effective scoring rate + advancement rate + offensive retention – turnover rate = total team points.
Now, because all of the stats used to create this model can be applied to an individual player as well as the team, I can also use the model to determine the value of a specific player on offence. What I have included here is the four factors applied to the players in the 2018 grand final and ranked those players in terms of offensive contributions only.
A couple of notes;
So I spent a bit more time on defence there than I really intended – but I will follow up with an indication of defence to fill the picture out in terms of best net overall performances a bit more clearly.
What I like about the results here in terms of offence is that the players who are valued highly are not necessarily high possession players. For me, this is interesting because I think it goes some way to helping understand scoreboard contribution in a more nuanced way.
Although we can always say that some of these players rely on ‘possession getters’ to set them up I think we can talk about those things separate from offensive efficiency – by developing a ‘facilitator metric’ for example or something to that effect.
I should also reiterate that each player’s result should be viewed as a total point contribution to the team’s final score. We are therefore saying that in the case of West Coast’s Josh Kennedy, for example, that he was the third most valuable offensive player on the ground and contributed slightly more than two goals worth of offence to a team who’s final score was only 79 points or 15.6 per cent of the team score.
Josh Kennedy of the Eagles. (Photo by Ryan Pierse/AFL Media/Getty Images)
Ultimately many of these highly rated offensive players will be levelled out by their negative defensive contributions and as such their net impact on the whole game will be more accurately reflected. From there we will be able to quantify things like ‘would the team have won without this player’ or even look at how altering a specific player’s performance would affect the result of the game.
For example a low possession player may be impactful offensively because they were efficient but this same player’s contribution will be more reactive to smaller shift in performance. Jaidyn Stephenson comes out as the most efficient offensive player here with a contribution of 13.85 points.
Some of his raw stats that contributed to this score included two goals on six kicks, three tackles inside 50 on 135 metres gained and two entries inside 50. If we adjust his numbers such that he sprayed one of shots at goal and had one less tackle inside 50 his contribution would drop all the way down to 7.8 points.
This would take him from the most efficient player on the field to the eighth and in theory cause Collingwood to lose by closer to 11 points as opposed to five.
For those who find this type of analysis illuminating I’ll add at the bottom here a four factors analysis and offensive ranking of the teams from 2018.
There is not a great deal of error here – which is good in terms of suggesting the model is accurate – however what we might be looking for is any disparities that might indicate a team under or over scored relative to their efficiency.
The Eagles, for example, scored 4.4 points more than their efficiency would predict. Arguably not a significant result. However if we factor in a standard deviation and the number of close games the Eagles played in 2018 it might be easy to imagine that a regression to their predicted output could change the result of two or three games.
Jack Darling and Josh Kennedy of the Eagles. (Photo by Will Russell/AFL Media/Getty Images)
Even this difference would drop West Coast from 16-6 and second position to 13-9 and eighth position.
Again, if we adjust just a couple of performance indicators we can see the difference we might expect in this particular teams output. West Coast were the fourth most accurate goal kicking side in 2018. Turning just one goal every other game into a miss instead would drop them to tie for 11th most accurate.
Couple this with a modest dip in metres gained per game of 200m and the predicted output for the team drops to 11th in the league from seventh (all else being held equal). For some these changes might not seem meaningful but it is a measurable way to understand the features of performance and the way that these elements impact the scoreboard.
Who’s in your ultimate team? Pick the best team and compete with other fans for daily prizes on Draftstars. For the best odds on the game try out Aussie bookmaker PlayUp. Think. Is this a bet you really want to place? Set a deposit limit.