For the past three years, HPN has attempted to project the strength of each side’s schedule in the upcoming season. With the 2019 fixture being finally released into the wild by the AFL (after a couple of delays), the time has come to look at the projected Strength of Schedule for each side in 2019.
What is Strength of Schedule?
In the AFL, the fixture since 2012 sees 18 teams play 22 games, which results in a slightly uneven draw. Teams face 12 opponents once and five opponents twice. A lot of attempts are made to quantify fixture fairness, and in this post we will deploy a measure called Strength of Schedule (SoS) to assess the difficulty of each team’s 2019 fixture.
SoS calculations can be used to assess the deviation of each team’s draw from a perfectly average opponent set. The concept is well developed in the USA due to a long history of mathematically assessing uneven fixtures. NCAA use SOS to determine entry to Bowl games and the College Football Playoff, the NFL use it to break ties for determining draft order. It is essentially the win-loss ratio of a team’s set of opponents.
We have calculated Strength of Schedule in three different ways:
- a basic win/loss balance;
- Pythagorean expectation based on points for/against; and
- Pythagorean expectation based on scoring shots for/against.
Originally attributable to the great Bill James and his work on baseball analytics, Pythagorean expectation has become a method of accounting for luck and telling us how many games a given team “should” have won. Matter of Stats has a thorough explainer on the subject of Pythagorean expectations and we use his 3.87 coefficient as calculated for the AFL.
Pythagorean expectation can be considered to represent expected wins, and tells us something about the underlying quality of each team in a season. They tell us in 2018 that due to scoring capacity or close losses that Melbourne and Geelong were probably better than elimination finalist teams on paper. Both were capable of scoring heavily, and Melbourne suffered a disproportionate number of close losses. Geelong, on the other hand, were often very good at restricting opponent scoring as well.
This method also tends to indicate how sides will perform in the coming season, and is the method HPN used in 2017 for winning The Arc’s finals prediction competition (along with the SoS calculations of draw). While the method is simple, it is solid too.
Beyond just raw points for and against, HPN also looks at scoring shots for and against, to further reduce variability for luck beyond results in close games. While different teams are experimenting with deliberately “low value” shots on goal, generally the side who generates more shots on goal is the superior one.
Other indications from the Pythagorean expectation are that Brisbane were also probably better than a five win side, and St Kilda also look like they could have won a couple more games, shaping more as a seven win side.
On the flip side, West Coast, Hawthorn and Richmond moderately outperformed Pythagorean expectations. However, it’s Sydney and Fremantle who were probably the teams who performed furthest above their underlying quality. Sydney fair particularly poorly on our scoring shots calculation due to conceding the most inaccurate scoring shots and kicking the most accurately themselves.
Overall draw difficulty
With the Pythagorean expectation in mind, and also using the tried and true counts of actual wins and losses, let’s move onto calculating the relative strengths of the 2019 fixture.
HPN has looked at the total opponent strength based on all three of the methods based above, and applied them to their 2019 opponent set. The league is sorted from the side with the hardest draw to that with the easiest (red is hard, green is easy):
On these predictions, Collingwood are shown as having the hardest fixture on an average of the three measures. The North and Hawthorn fixtures look slightly harder on the Pythagorean expected win/loss of their opponents, while Melbourne’s opponents had a slightly better actual 2018 win/loss record.
Richmond show as having the fifth easiest draw, and that points us to an important effect on draw difficulty – one’s own strength. Teams do not play themselves, so all things being equal, the worst team should have the hardest opponent SoS in an even fixture.
How well is the draw handicapped?
The AFL fixture is inherently uneven, and as a matter of specific policy it uses the balance of doubled opponents to compensate for team strength. Like the best racehorse in the race carrying extra weight, the top teams are handicapped by being given more difficult fixture burdens. The top six teams (post-finals) are supposed to play more fellow top teams, and the bottom teams play each other more.
HPN has calculated the strength of the rest of the league for each side, and shown how different to this ideal each actual draw is. If teams played every team once or twice, the middle column is how their Strength of Schedule would look.
Richmond’s fixturing is roughly line with the AFL’s policy. Although they’re luckier than most of their fellow top six sides, the Tigers’ apparently easy fixture is based mostly on them having been the best side through the regular season last year. A weighted fixture can only do so much to overcome a team’s own strength, but the Tigers also avoid some of the toughest double-ups in each of the three brackets.
At bottom end of the ladder, it’s clear how strongly the fixture handicapping has overcome the naturally more difficult fair draws that those sides would face. The bottom six have been well protected.
GWS and Adelaide are the teams who can consider themselves the luckiest, with Adelaide fixtured nearly as favourably as a bottom six side, while GWS sit below two middle bracket teams in Essendon and North Melbourne on opponent difficulty.
Expected wins from fixture
So far we’ve been talking in win-loss percentages, but we can make things a little more concrete by converting the hypothetical win-loss records into actual expected wins above or below the fair fixture. This involves the complex mathematical procedure of multiplying the figures above by 22 (number of games per season) and deducting the product from 11 (the number of wins per team given the number of games).
This chart shows, for example, that if an average team played an average set of non-St Kilda teams, they’d be expected to win about 0.8 less games than if they played against St Kilda’s actual opponent set.
For 2019, no fixture is an entire expected win away from the fair fixture ideal. At the difficult end, Collingwood is starting next season just over half a win behind the 8-ball, with St Kilda getting a 0.8 win head start. That means that overall, there is about a 1.4 win gap between the toughest and softest opponent sets, compared to the fair fixture which sits somewhat in between them.
Looking back: projections won’t match reality
One thing we can say with certainty is that the projected Strengths of Schedule will not pan out quite like presented here. Heading into 2018, nobody would have suspected that St Kilda and Sydney faced the hardest draws, but the improvement of sides like West Coast, North Melbourne, Hawthorn and Melbourne made this so.
Here’s what actual in-season results did to the projected draw difficulties, again in expected wins:
St Kilda turned out to have a fixture 0.8 wins more difficult than a fair one (accounting for their own strength and not playing themselves) while Sydney also fared unexpectedly badly. The Saints saw the biggest change in the projected strength of their draw, as it got a full expected win worse than expected.
At the other end of the scale, North Melbourne’s fixture turned out to be 0.9 wins better than the fair ideal, and Melbourne’s draw opened up by nearly one expected win due to the collapse of Adelaide, the decline of St Kilda and the more modest slides of Geelong and the Western Bulldogs.
Fixture impact of 2018 compared to 2019
By comparing how far from the “fair draw” teams were in 2018 and how far their 2019 fixture is from fair, we can measure how much teams should expect their results to shift purely due to fixture impact.
The biggest winner here is clearly St Kilda. Saddled with a surprisingly hard 2018 draw, they’ve been given a soft fixture worthy of their lowly ladder position. The cumulative impact of doubling up on GWS, Hawthorn, Melbourne, North Melbourne and Richmond in 2018 and then in 2019 instead facing Carlton, Gold Coast, Fremantle, Adelaide and Melbourne, is about 1.5 wins. Unless a couple of those sides jump up the ladder, of course.
Other winners compared to 2018 are the Bulldogs, Sydney, Adelaide and Gold Coast to the tune of about half an expected win each.
Teams who should expect a rougher time are primarily North Melbourne, Melbourne and Collingwood; all teams who got lucky with their 2018 opponents, who jumped up a bracket or two relative to their 2018 season. They’ve all been given fixtures about a win harder than in 2018. West Coast and Hawthorn, also teams who moved up a bracket in 2018, see more modest increases in fixture difficulty.
What this all means
If you want to reduce everything above into some neat talking points, well here they are:
- Collingwood have the hardest opponent set overall, with Melbourne and Hawthorn’s also very difficult. St Kilda or Gold Coast have the easiest.
- The fixtures of Collingwood, North Melbourne and Melbourne for 2019 are all about one expected win harder than in 2018, while St Kilda’s is 1.5 wins easier.
- Melbourne, Geelong, Brisbane and St Kilda probably underperformed in 2018 relative to their Pythagorean expectations, while Sydney and Fremantle look like they were over performers.
- The fixture effectively handicaps the top six sides and helps the bottom six sides as intended, but Adelaide and GWS appear to have slightly weaker than expected opponent sets.
- In 2018, St Kilda and Sydney had fixtures turn out a lot harder than projected using this method in preseason, while North Melbourne, Melbourne and West Coast saw their draw become easier as doubled-up opponents declined in strength.