Getting it wrong

6176178769_f9dc50d7cd_z

Introduction

Each week, I try to predict the outcomes of the nine AFL games played here in Australia.

I lack the detailed insights and understandings that Tony Corke, Darren O’Shaughnessy and Sam Robertson, amongst others, bring to their analysis of football.

I do not watch any games either so it is prediction process that is entirely dependent on my interpretation of three indicators:

  • A team’s final regular season ranking position in 2015.
  • A team’s current position on the AFL ladder.
  • The betting odds for a win for each team.

In Round 7, I managed a 7:2 win:loss outcome. My biggest error was to tip the Gold Coast to beat Melbourne. Two of my three indicators suggested a Melbourne win. I was swayed by a small margin in the odds in favour of Gold Coast and their home ‘advantage’.

At half time, I thought I was doing well, it was a contested half of football with Melbourne leading by 5 points (I am happy if a gap is + or – one goal):

GCMHT

Source: AFL Match Centre

Thereafter, Melbourne scored nine goals and four behinds (58 points) in the third quarter to Gold Coast’s three goals and one behind (19 points).

The game of two halves looked like this as Melbourne added 47 points in the fourth quarter (to Gold Coast’s 18):

GCM7

Melbourne scored two goals in the first three minutes of the third quarter and then scored three more before the Suns first goal of the quarter after 11 minutes.

Melbourne’s third quarter performance was their best since 1994.

14140701429_0266be525d_z

Decision Support

One of my hopes in monitoring performance is to contemplate how analysts might support coaches.

Even after fourteen years in Australia, I am still perplexed by the idea of a coaches’ box at football games. Whenever I catch glimpses of coaches in these boxes and see the volume of people and computers in the boxes, I wonder how the available data is processed, shared and acted upon.

In the Gold Coast game, I would have mentioned that five games have been won this season by the team losing at half time. Apart from Port Adelaide’s win in Round 1 after being 6 points down and then winning by 33, all the other games have been single figure recoveries. My big message to share with coaches was the precision of the start to the third quarter. I would be keen to explore the active rest possibilities at half time too, in order to ensure our start to the third quarter brought us back into the game.

I would be mindful as a Gold Coast support staff that the game got away from us rapidly in Round 6 and I would like this to be my anchor for the technical and tactical response to Melbourne on our home ground in front of our own supporters.

Round 6 v Geelong:

Round 6

The AFL feed from the Round 7 game reported:

3q

This is the tipping point in the game in my secondary analysis. Only one team, ironically Melbourne in Round 1, has overcome a 20 point deficit at the end of the third quarter to win a game this season.

My hope at this point in the game is that the scenario-based training we planned leading into the game helps us to claw our way back and set up a big final quarter. I would have drawn upon our experiences this season against Fremantle and Brisbane to provide some evidence based practice of what we can achieve.

So whenever I do get a prediction radically wrong, it turns out to be a powerful coach learning opportunity and a way for me to contemplate how I might be of service as a backroom staff member.

I am optimistic enough to believe that getting it badly wrong can enhance the process of getting it right.

Photo Credits

The Mark (Drew Douglas, CC BY-NC2.0)

Run on (Petra Bensted, CC BY 2.0)

Postscript

A comment by Graham on this post has led me to Ryan Buckland‘s discussions of AFL in The Roar. (“As an economist, Ryan seeks fix the world’s economic troubles one graph at a time. As a sports fan, he’s always looking one or two layers beneath the surface to search for meaning, on and off the field”.)

6 thoughts on “Getting it wrong”

  1. Excellent comments there on how analysts might support coaches. I talked to the NRL analysts last week and mentioned how many screen designs for in-play are lit up like Xmas trees in green and red for every little indicator that is going astray. This just turns a wealth of data into … an array of colourful data.

    You really need to design an algorithm that just filters the most significant numbers (those that are causing the most damage) to the top and colour-codes them according to significance, both statistical and coach-specific. Also consider which KPIs are “coachable”, and whether scenario planning has covered how to respond. Very difficult for even the best numerate coach to put in place a spontaneous plan just in response to a pattern of numbers. A qualitative layer is essential.

    Thanks again Keith

    1. Thank you for finding the post, Darren.

      I think the qualitative understanding is the synapse between the back office and front office.

      As I was reading your comment I was thinking about the illuminations on share stock screens.

      I keep hoping the qualitative work makes IF…THEN actions possible. In the Gold Coast game there were six minutes in which to intervene … a player-led response in the field of play.

  2. Hi Keith, another writer who produces insightful analysis of AFL games is Ryan Buckland ( http://www.theroar.com.au/author/ryanbuckland7/ ) . Ryan’s background is economics, and he brings a qualitative approach to his analysis. In particular, I like the way that he recognises the uncertainty / randomness in all the various statistics that are produced about AFL (and all sports).

    1. Hello, Graham

      Thank you for finding the post and suggesting Ryan. One of the joys of open sharing is that we can learn continuously. 🙂

      Best wishes

      Keith

Leave a Reply