So with a new season defined we can being to analyse fixtures. This is my first one so I'll pretty much just explain how I came up with my assessment and sum up the results in a few charts.
First, the methodolody.
I considered 4 hypothetical seasons:
- Each team faces an average opponent every week at a neutral venue.
- Each team faces an average opponent (exlcuding themself) every week at a neutral venue.
- Each team faces their fixtured opponents, but always at a neutral venue.
- Each team faces their fixtured opponents at the fixtured venue.
I still have to write a blog explaining how my ELO model determines strength of opponent and home ground advantage. At a high level though, team strength is based on shots taken and shots allowed. Home ground advantage is based mainly on the travel distance, with a bit of historical team-venue performance considered too.
With these 4 hypothetical seasons we can then compare the expected wins or total expected margin for each team to see the contribution of different factors in each teams strength of schedule.
- 1 vs 2 tells us the effect of teams not having to play themselves. Good teams benefit from this while poor teams are disadvantaged.
- 2 vs 3 tells us the effect of each teams double up opponents. As each team only gets to play 5 opponents twice, we need to account for the fact that each team will not play against the real average available to play against.
- 3 vs 4 tells us the effect of home ground advantage (HGA).
With the AFLs fixturing rules, teams in a given bracket of 6 will get more double ups against teams in the same bracket than teams in the others. This should show itself in the second comparison (2 vs 3) listed above.
The first comparison (1 vs 2) by definition will show us that good teams benefit form not playing themselves while poor teams are disadvantaged by not getting to play themselves.
It will be interesting to see what the 3rd comparisons reveals...
First I've evaluated the strength of schedule using difference in expected wins. Difference in expected wins represents how many extra or fewer wins I expect each team to finish the season on due to each fixture effect (No playing self, opponent strength and HGA).
The chart below shows each team on the x-axis and the difference in expected wins due to each fixture effect on the left y-axis as stacked bars.
I've plotted teams total difference in expected wins as a line, and the strength of each team according to my model as a dot. Here the team strength represents the probability of that team beating an average opponent at a neutral venue, so it's between 0 and 1.
These charts are interactive so you're able to toggle each series on and off to focus on just a few things at once.
We can see Collingwood faces the toughest draw and St Kila the easiest in this representation. By far the biggest contributor is the opponents played. The green bars tend to dwarf the blue and black ones.
The problem with looking at difference in expected wins though, is that expected wins show diminishing returns to any advantage.
For example when we look at the home ground advantage factor for Fremantle and West Coast, they're pretty different (0.64 wins compared to 0.05 wins respectively), despite them both having pretty similar travel schedules for themselves and their opponents:
- Fremantle host 1 extra VIC team compared to West Coast while West Coast host 1 extra SA team.
- West Coast travel to Melbourne 1 more time than Fremantle while Fremantle travel to Adelaide 1 more time than West Coast.
The reason is that we already expect the Eagles to be good, and so the boost they receive from HGA doesn't have the same effect on their win probability as it would for a poorer team like Fremantle.
Because of this, I decided to look at differnce in expected margin across the season too. Difference in expected margin represents how many extra or fewer points I expect each team to accumulate throughout the season due to each fixture effect (No playing self, opponent strength and HGA).
The left y-axis here represents difference in expected margin due to each fixture effect. The rest of the chart is the same as above.
Here West Coast and Fremantle have much closer HGA components. Non VIC teams tend to have the highest HGA components too, with the exceptions of GWS and Sydney...
Interestingly, St Kilda has one of the highest HGA components. I've looked into this and there are a few things at play here.
- St Kilda gets to play teams that travel further than St Kilda has to travel
(Eg they play Brisbane and West Coast at home and Sydney and GWS (Canberra) away)
- St Kilda gets to play at neutralish venues or venues where their opponent also has to travel,
negating some or all of their opponents HGA.
(Eg GCS @ Townsville, Port @ China, North @ Tasmania)
Overall, both representations show similar results:
- Stronger teams tend to have harder fixtures.
- Collingwood, Melbourne and North Melbourne have the hardest fixtures.
- St Kilda has the easiest fixture.
- Adelaide got a good fixture given their team strength
- Richmond for a harder fixture than Adeliade, but still quite a bit easier than they should have.
I'm also impressed that over the season, the biggest absolute advantage or disadvantage sits at about 3 points per game, or about 5% in win probability per game.
I had no expectation prior about what an appropriate advantage would be, but 3 points/5% seems pretty small for the maximum. I'm sure there are plenty of other more important factors at play in determining outcomes for teams over the season, such as:
- Injuries (both in season and in game)
- Luck in facing teams at the right/wrong time
- Off season recruitment and/or game plan changes
- Finding new rule change loopholes and exploiting them before other teams do...
Anyway that's it for my fixture analysis, hope you got something out of it!