MAT4MDS Comprehension task, Summer, 2024/2025

Submit by 11:59 pm on Tuesday, February 4 (Week 11).

Dominance matrices in AFL Do they actually work?

Objectives
The method of dominance matrices is a mathematical tool that allows for the ranking of teams in a
round-robin tournament, even when not all teams have played each other. Before beginning this project,
it is recommended that you read Roger Walter’s article, “Dominance Matrices: Your Secret Weapon
in Footy Tipping,” watch the three videos about dominance matrices available in the section ’Final
mathematical comprehension and explanation task, and review your answer to the initial task and the
solutions provided. The purpose of this project is to determine the reliability of the dominance matrix
method using historical AFL data, which is freely available online, e.g. on www.footywire.com.

Detailed project description

Historical AFL data is obtained from www.footywire.com. Consider the 2015 season as an example. In
this year the Hawthorn Hawks (Hawks) beat the West Coast Eagles (Eagles). The AFL footy ladder
after all 23 rounds is given in Fig. 1.

2015 AFL Season Ladder sorted by Premiership Points

Position Team Played Win Loss Draw %Won Points For Against Percentage Movement Streak

1 Fremantle Dockers 22 17 5 0 77.27% 68 1857 1564 118.73% +0 Lost 1

2 West Coast Eagles 22 16 5 1 75.00% 66 2330 1572 148.22% +0 Won 1

3 Hawthorn Hawks 22 16 6 0 72.73% 64 2452 1548 158.40% +0 Won 2

4 Sydney Swans 22 16 6 0 72.73% 64 2006 1578 127.12% +0 Won 4

5 Richmond Tigers 22 15 7 0 68.18% 60 1930 1568 123.09% +0 Won 4

6 Western Bulldogs 22 14 8 0 63.64% 56 2101 1825 115.12% +0 Lost 1

7 Adelaide Crows 21 13 8 0 61.90% 54 2107 1821 115.71% +0 Lost 1

8 North Melbourne Kangaroos 22 13 9 0 59.09% 52 2062 1937 106.45% +0 Lost 2

9 Port Adelaide Power 22 12 10 0 54.55% 48 2002 1874 106.83% +0 Won 4

10 Geelong Cats 21 11 9 1 54.76% 48 1853 1833 101.09% +1 Won 1

11 GWS Giants 22 11 11 0 50.00% 44 1872 1891 99.00% -1 Lost 1

12 Collingwood Magpies 22 10 12 0 45.45% 40 1972 1856 106.25% +0 Lost 1

13 Melbourne Demons 22 7 15 0 31.82% 28 1573 2044 76.96% +1 Won 1

14 St Kilda Saints 22 6 15 1 29.55% 26 1695 2162 78.40% -1 Lost 2

15 Essendon Bombers 22 6 16 0 27.27% 24 1580 2134 74.04% +0 Won 1

16 Gold Coast Suns 22 4 17 1 20.45% 18 1633 2240 72.90% +0 Lost 2

17 Brisbane Lions 22 4 18 0 18.18% 16 1557 2306 67.52% +1 Won 1

18 Carlton Blues 22 4 18 0 18.18% 16 1525 2354 64.78% -1 Lost 2

Figure 1: 2015 AFL Season Ladder sorted by Premiership Points.

All 2015 AFL fixtures with the final score for each match can be found at https://www.footywire.com/
afl/footy/ft_match_list?year=2015, and also on the LMS site.
Our goal is to see if the dominance matrix method applied to the 2015 season after a certain number of
rounds m < 23 will give similar team rankings, as the actual AFT footy ladder after 23 rounds in Fig. 1.
In these 23 rounds each out of 18 teams played 23 games, some teams played each other twice.

By preparing your report proceed as follows:

ˆ Choose several values of m (at least three), such that 10 ≤ m ≤ 22. For example, m = 10, 14, 18, 20.
ˆ Use the data in

AFL Fixture 2015.pdf

to create an 18 × 18 dominance matrix D(m) for each value of m that you have chosen. For
constructing dominance matrices D(m) you have a choice between the win-loss records method
described on page 8 by Roger Walter article or the ratio-scores method described on page 9.
To construct D(m) use all the results of each game played up to round m (including round m,
but NOT only round m).
If you choose the win-loss method, give 4 points for a win, 0 point for a loss and 2 points
for a draw (this is standard practice for footy ladders).
When creating the dominance matrix, use the same order of teams as in the actual AFL ladder in
Fig. 1. In other words, Fremantle Dockers is team 1, West Coast Eagles is team 2, Hawthorn Hawks
is team 3, etc.
In this way, for example for m = 10, for ratio-scores matrix D(10) you will have D(10)3,10 =
123/61 and D(10)10,3 = 61/123, because Hawthorn Hawks defeated Geelong Cats 123 − 61 in
round 1. On the other hand, D(10)10,13 = D(10)13,10 = 0, because Geelong Cats played against
Melbourne Demons only in round 12, and, therefore, this result was not available after round 10,
when the dominance matrix is constructed.
In situations where two teams have played with each other twice within a given m, consider the
most suitable approach for incorporating score ratios or win-lose points into the matrix D(m). You
may explore different options, such as solely considering the last result, calculating the average
between the two results, or experimenting with your own innovative ideas.
ˆ After the dominance matrix D(m) is constructed for the chosen values of m, calculate (D(m))2 .
Then, calculate the predicted team rankings using the formulae

r = r1 + 0.5r2 ,

where the first order ranking r1 is

r1 = D(m)1 − (D(m))T 1

and the second-order ranking r2 is

r2 = (D(m))2 1 − ((D(m))2 )T 1.

Explain in your own words the meaning of r1 , D(m)2 , and r2 .

ˆ Compare the predicted team rankings r with the actual AFL ladder after 23 rounds in Fig. 1. Which
teams have been ranked correctly? Discuss the results. Can the results be logically explained?
 ˆ What is in your case the smallest value of m such that the dominance matrix method gives the
correct prediction for the most number of teams.
ˆ Based on your results comment on the reliability of the considered dominance matrix method.
Would you trust this method early in the season?

ˆ Attempt to determine the best possible formula for predicting the true ranking for the end of the
season. In doing so consider experimenting with the following parameters:
- changing the weights for the first and the second order rankings r1 and r2 ;
- extending the formula by adding higher order rankings;
You are also welcome to use your owns ideas for finding the best possible formula for predicting the
true ranking.

ˆ If you work in a team of two or three, conduct the analysis for both methods described above of
constructing dominance matrices. Compare and discuss the obtained results.

Your report is to be prepared and presented as if you are a consultant. This means that you must
present your findings in as persuasive a way as possible, explaining the reasons for every technical choice
you made. There may be some choices that are good under different circumstances, but bad for this
circumstance (or vice versa). You should choose between them and explain your choice. At the start of
your presentation you must explain the idea of the method including the meaning of r1 , D(m)2 and r2 .
(Dominance matrix D(m) should be demonstrated within the presentation for at least one value of m).

There are two parts to the assessment submission:

Part A: You can work individually or in group of either two or three to prepare a “PowerPoint” style
report. This should be submitted as a pdf of up to 6 pages/slides (1 person)+title page, 8 pages/slides (2
people)+title page or 10 pages/slides (3 people)+title page. An overall higher standard will be expected
from larger groups than individuals. The first slide must include the names of group members
(this slide can also include analysis information).

Part B: You must work individually to prepare a video in which you explain your (or your group’s)
predictions and choices. You should record yourself talking over the screen share of your report using
zoom or a similar tool. The video recording must show you. This recording must be no longer than 5
minutes. The report you screen share must match your submitted pdf.
The video recording must show you at the start. You have to introduce yourself and show
your id card (Student Id or Driving license or similar) to the camera, so that it will be
possible to recognise your photo and your Name on the card.
If you do the analysis and slides in a group, each group member is required to submit a presentation
and a video individually, presenting and discussing the results. You must mention the names of all group
members at the beginning of the video. The project will be graded individually.

The marking guide to this task is available on LMS in the section Final mathematical comprehension
and explanation task.