Dogs, Tigers, Medians and Moments: Reflecting on the 2017 #AFLGF


Richmond Tigers have won the 2017 AFL Grand Final. They defeated Adelaide by forty-eight points.

Two days ago, I used median scoring profiles to suggest, all other things being equal, an Adelaide win. My season’s data for the competition gave me these end of quarter scores:

In my post, I suggested that the challenges to an Adelaide victory were:

  • The closeness of the contest in the first quarter of the game
  • Whether Richmond stayed with Adelaide in the 2Q
  • If Richmond lifted in 3Q

Should Richmond mange these three conditions, I thought this would mean it would come down to which team had the legs and tactical nous in 4Q.

The scores by quarter were:

Richmond turned an 11 point deficit at the end of the first quarter into a nine point lead at half time. The end of the first quarter was the high water mark for Adelaide in the Final. Richmond outscored Adelaide in the second quarter by 20 points, lifted in the third quarter to outscore Adelaide by 25 points, and had the legs in the final quarter to outscore Adelaide by 14 points.

The Dog in the Night-Time

My all-things-being-equal-model indicated an Adelaide win. I was mindful that Richmond’s opportunities to own the game required some conditions to be met. I identified these conditions using median scoring profiles for each quarter of games played over the whole season for both teams.

The result of the game sent me back to a paper written by George Lewis and Jonathan Lewis in 1980, The Dog in the Night-Time: Negative Evidence in Social Research. They proposed:

In social research, there is an overwhelming emphasis upon collecting positive data, whether it be in the form of statistically significant attitudes, important documents, or observer descriptions of unique settings. This emphasis, while responsible for shedding much light in previously dark areas, none the less has had the important and dangerous side-effect of minimizing the worth of negative evidence, that is, the significance of a thing’s absence. (1980:544)

Their paper suggests a paradigm of seven types of negative evidence:

(1) Events Do Not Occur; (2) Population Is Not Aware of Events; (3)Population Wishes to Hide Events; (4) Commonplace Events Are Overlooked; (5) Effects of the Researcher’s Idea Set; (6) Unconscious Non-Reportage; and (7) Conscious Non-Reportage. (1980:544)

They conclude with the observation “in order to gain the best perspective on the problems one is studying, there must be a greater emphasis on uncoverlng and utlilzlng negatlve evidence” (1980:555).


This post is my way of responding to negative evidence and is an exploration of the phenomenographic reporting of the game. I did not watch the game but monitored the score on the AFL web site.

Moments in the Grand Final for me were:

Post game I looked at the official AFL match feed and picked out these moments:

This led me to think about the conversations both coaches would have had with their respective teams at half time. I felt that a nine point half-time difference would make the third quarter of the game a fascinating contest of wills and dynamical tactical responses.

Note the word ‘obliterated’ in this Q3 summary.

The outcome of these moments was Richmond’s first championship in 37 years. Now that this game is in the data base we have an opportunity to contemplate how the Final might have been played differently.

Moving from If …Then to Yes … And

I confess to not going out in my back yard to replay the moments of the Grand Final. I did imagine how these games might be played out in yards in Richmond and Adelaide.

What the negative evidence of this Grand Final has encouraged me to think about is how we might structure learning environments. I am more and more convinced that if we were to move away from prescriptive IF … THEN plans to a much more improvised YES … AND encouragement, we might be able to be even more responsive to changes in game state and tactical dispositions.

YES … AND accepts negative (and positive) behaviours and dampens (or amplifies) them. In my data terms, this would enable teams to be true to type in any game context … even with the aura of playing in a Grand Final.


Photo Credits

We’ve done it (Richmond FC, Twitter)

Well that was fun (Caitlin Arnold, Twitter)

Profiling the 2017 #AFLGF Teams


I have kept a record of each quarter played in this year’s AFL competition. My data are compiled manually from the official AFL site.

On the eve of the Grand Final, I have explored the median profiles of both teams in the Final, Adelaide and Richmond. The odds offered by one bookmaker on Friday, 29 September were: Adelaide $1.73, Richmond $2.15.

Darren O’Shaughnessy puts the likelihood of Adelaide winning the AFL Grand Final at 59.1%. Tony Corke provides an overview of predictions for the Final and concludes “Adelaide are forecast to score between 86 and 90 points, and Richmond between 82 and 88 points”.

The Teams

My data are presented here as box plots generated with the BoxPlotR web-tool:



Adelaide Richmond Direct Comparison


Although I have kept a record of every quarter played this season in the AFL, I have not watched one minute of play. I do not read any of the newspaper coverage of the game nor do I listen to any of the football programs during the season.

I hope to find patterns in AFL scoring data and use only one indicator, the score at the end of each quarter of the game. For Adelaide and Richmond, these data give me these median profiles for each quarter of the games played. (Note that I have corrected these data 30 September.)


In terms of “all other things being equal“, this gives me a three goal advantage to Adelaide (18 points) (Note I revised this figure pre-game after reviewing my data). I will not be watching the Grand Final but, post event, I will be interested to learn about:

  • The contest in the 1Q
  • Whether Richmond stayed with Adelaide in the 2Q
  • If Richmond lifted in 3Q

All of which would mean it would come down to which team had the legs and tactical nous in 4Q.

Otherwise we do have a predictable outcome (other things being equal), an Adelaide win. A win for Richmond will lead to fascinating conversations about readiness to perform.

Whoever wins, the day for me, in large part, will be spent remembering Phil Walsh.

Photo Credits

Grand Final Parade (AFL, Twitter)

Phil Walsh (Wikipedia, Fair Use)

Episodes in a contested AFL game: an example from Round 7 in 2016



Yesterday, I discussed the Gold Coast versus Melbourne AFL game from Round 7 of the 2016 AFL season.

Today I am looking at a game in Round 7 that was contested over three of the four quarters of the game, Richmond versus Hawthorn.

I had anticipated that Hawthorn would win this game.

This is the scoring chart for the game from the AFL website:


A Contest

I think this is an excellent example of a contested game in which momentum shifts from one team to another until in the fourth quarter, Hawthorn breaks away and dominates the game.

In each of the first, second and third quarters, Richmond leads the game:

  • Start to 13.11 (1Q)
  • 11.03 to 21.41 (2Q)
  • 12.34 to 19.57  (3Q)

Hawthorn’s responses came:

  • 13.12 to 21.13 (1Q)
  • 21.42 to 30.28 (2Q)
  • 19.58 to 29.36 (3Q)


Momentum Shift

There were nine points between the teams going into the final quarter (Richmond 71, Hawthorn 80).

Richmond is one of the four teams who had overcome a nine point or greater deficit at three quarter time to win this season (in Round 1). Ironically, Hawthorn has been involved in two games when they have come from behind at the end of the third quarter (Rounds 3 and 5). Hawthorn has been tied at the end of the third quarter too (Round 4).

Hawthorn dominated the first 11.44 minutes of the final quarter and then closed out the game from 19.10 onwards.

As with yesterday, I am interested in the interaction between on-field performance and off-field coach observations.

The AFL text commentary noted in the third quarter:


and later in the quarter:


When Hawthorn scored a behind at 6.57 in the final quarter, I identified this as a tipping point in the game. Hawthorn now lead by 22 points and it is a moment of no return for Richmond. They must respond. They do equal Hawthorn’s scoring over the next 12 minutes (1 goal and 1 behind each) but are then overwhelmed by four Hawthorn goals in five minutes. In this phase of the game Hawthorn amplified their momentum and Richmond were unable to dampen it.


A game theoretic approach to performance

My interest in observing AFL is part of my fascination with finding ways to share stories of performance in real-time and lapsed-time.

It would have been interesting to hear the end of third quarter conversations between coaches and players for both teams. It would have been informative to hear player talk in the first eight minutes of the fourth quarter too.

These voices would form part of what Philippe Mongin (2009) calls A Game-Theoretic Analysis . In that paper he writes:

military campaigns provide an opportunity for successful application of the formal theories of rational choice. Generalizing the argument, we finally investigate the conflict between narrative accounts – the historians’ standard mode of expression – and mathematical modeling.

This seems a potentially rich way to explore performance.

In the post match conference, the Hawthorn coach is quoted:

We started to win the ball a little bit better from clearances and, in the early part of the game in particular, Richmond were really strong out of the centre bounce. In the last part of the game that started to flow our way. We started to think if we could get enough supply we could kick a big enough score to win, but we didn’t anticipate that it was going to open up like that.

The Richmond coach:

We were disappointed to go down like we did in the last quarter. We were just belted around in clearances and contested ball that last quarter and they just controlled the ball. The first three quarters I thought were commendable. Our effort and intensity was there and we played some pretty good footy. It was probably a game where we had a couple of opportunities during the third (quarter). I thought we didn’t capitalise enough. (We) just had some bits of play where Hawthorn were too classy and made sound decisions, whereas we probably coughed it up and turned the ball over.

Both coaches’ observations raise some fascinating issues around player-led on-field initiatives and decision-support conversations in the coaches’ box.

Contested games give us an opportunity to travel between the stories coaches’ construct and the data that informs their construction.

A great round of AFL for conversations about performance.

Photo Credits

Cats v Hawks (James D Photography, CC BY-NC-ND 2.0)

Richmond vs Melbourne rnd4 (Rachel Hofton, CC BY 2.0)

Melbourne Cricket Ground 1870-75 (Robert Smith, CC BY-NC-ND 2.0)