Wednesday, February 18, 2015

Revisiting Home Ground Advantages in the AFL

How much is home ground advantage in the AFL worth?

In my AFL Power Rankings I adjust the final margin of a match by home ground advantage before allocating ranking points to teams. For example, when Sydney beat Fremantle by 24 points in their qualifying final last year, I subtracted 12 points from the margin based on Sydney having a two-goal home ground advantage (before then adjusting the margin for the strength of the opposition).

I have also used these home ground advantage (HGA) adjustments over the past few years to rank the difficulty of each AFL team’s fixture.

In August last year on this blog, in a response to a question I posed about North Melbourne’s HGA in Tasmania, some comments were made by Manikato888 regarding HGA in the AFL. Since then we have been chatting back and forth about what may be good HGA adjustments to use. Manikato888 was kind enough to share with me the work he had done on which adjustments seemed to work best in predicting the results of matches.

These conversations have led to me revisiting the HGA adjustments that I use. I haven’t changed them completely to match the adjustments that Manikato888 uses, in part because where the differences were very small I wanted to maintain consistency wherever possible with what I had used in the past. But I have made changes in cases where both logic and empirics suggested a slight change may be warranted.

For example, I had different adjustments for the Swans and GWS against Victorian teams, with the Swans having a six point HGA against Victoria teams in Sydney, while the HGA for GWS was 12 points. Since they are both from the same state it made sense to me to equalise their HGAs, even if we do not have a lot of data to judge the HGAs of GWS yet.

Another change I made was to reduce by a goal Geelong’s HGA in Geelong against other Victorian teams. This makes it no more than Sydney’s HGA against Victorian teams which made sense to me given Geelong is a hell of a lot closer to Melbourne than Sydney is. For games between Geelong and Victorian teams in Melbourne I am keeping the HGA as neutral – i.e. zero. I considered giving the Melbourne teams a goal HGA, but decided against it given that the Cats play a fair amount of home games against Victorian teams in Melbourne.

I have also reduced by a goal Hawthorn’s HGA in Tasmania against non-Victorian teams. I previously had this as the same HGA – that is, two goals – as if they played in Melbourne. Based on empirics, a case could have been made for reducing this further to zero, but given the Hawks have played a few games in Tasmania per year for some time now I wanted to maintain some HGA.

The table below summarises these changes:

Home Ground Advantage
Old
New
QLD v non QLD
12 pts
12
SA v non SA
12
12
WA v non WA
12
12
NSW v non NSW and non VIC
12
12
VIC v non VIC
12
12
Swans v VIC & VIC v Swans
6
6
GWS v VIC & VIC v GWS
12
6
Cats v other VIC in Geelong
12
6
Hawks v non VIC in Tasmania
12
6
Everything else
0
0

Keep in mind with all this HGA adjustments are relatively unimportant in terms of determining the rankings. Generally, over the course of a season, and even over the course of a few weeks, the HGA adjustments for a team will net out to zero or close to it. Also, the most weight that any particular game carries in the rankings (which is the most recent game) is less than nine per cent. Hence, a HGA adjustment of 12 points in a particular game would only affect a team’s ranking by at most one ranking point.
 
Given that they have a small effect I am only going to adjust the HGAs going forward rather than going back and changing my historical rankings. For the most part these changes will be imperceptible … but they’ll make me feel a little better about my rankings (I think).

No comments: