The answer is Geelong, closely followed by Townsville:

Did you want some more explanation or the graph is enough? The graph is fine? Well, I’ll just sit here in case you change your mind. Here’s an appropriate moment from The Simpsons:

I can’t take the silence anymore so I’m going to keep talking. Some points about the graph above:

1. Population for regional centres is only given every five years in the ABS dataset I’ve been using as the basis for Cap calculations, so we only get limited snapshots in time.
2. The Gold Coast didn’t have a team between 1999, when the Chargers folded, and 2007, when the Titans were snuck in the backdoor. This is the same market that will apparently support the Suns. The Cowboys only joined the ARL in 1996, so there’s no datapoint for Townsville in 1991.
3. Small town teams are popular, really popular. Geelong’s Cap is better than that even of Melbourne. I reckon if we had data for 2015, the Cowboys’ Cap would be around that of Geelong’s 2011 premiership year.
4. Even mid-sized cities have better Caps than their bigger cousins. The Gold Coast’s Cap, where it would seem locals have better things to do than watch the Titans get thrashed, is better than Brisbane’s has ever been. Newcastle are firmly behind the Knights, even though the 2016 team was objectively the worst fielded in the NRL’s history and they haven’t won anything in 15 years. Canberra’s lighter support for the Raiders is perhaps explainable by the number of ex-pat Victorians that live in the nation’s capital.
5. In big cities, Cap decreases when a new team is added. Compare Perth’s Cap from 1991 to 1996 (Fremantle’s first season was 1995) and Adelaide’s Cap from 1996 to 2001 (Port Adelaide’s first season was 1997).

If we put aside my Newcastle and Gold Coast snark, there’s some interesting threads to examine there.

Point 4 suggests a relationship between the population of the city and Cap. I’ve scatter plotted the 1991, 1996, 2001, 2006 and 2011 Caps for Townsville, Gold Coast, Canberra, Newcastle, Brisbane and Sydney ruby league sides and the Perth, Adelaide, Geelong and Melbourne AFL sides:

A R-squared of 0.22 suggests there’s a degree of correlation. The interesting thing is that the polynomial trendline offers a best fit out of Google Sheets selection. Basically, small towns really love their teams, perhaps because there’s not much else going on, and Sydney and Melbourne love their eight-plus teams, because of a century of tradition, but the mid-sized population centres which have one or two teams – from Canberra through to Brisbane – are a bit meh by comparison.

What if we tried plotting Cap for those same data but adjusting the population for the number of local teams:

The linear trendline has an R-squared of 0.58 (removing the Broncos’ outlying data yields a R-square of 0.2), indicating a weak but definite correlation between Cap and the number of teams available to a given population. More teams per capita correlates to a better total attendance. 

This is somewhat intuitive. The more teams there are means more games to attend, and so a greater total attendance, and a cross-town rivalry can help generate interest locally. There’s obviously an upper limit – you couldn’t just add three dozen teams to a city and watch Cap drop away to zero – but we lack the data at the NRL/AFL level where a mid-sized city has expanded from two to three or four teams to see what happens. If the A League gets a third Sydney or Melbourne team, this might be instructive.

What this doesn’t mean is a better average attendance. Let’s say Brisbane expands from one to two teams. Based on the Adelaide and Perth experiences, Cap might drop by a factor of one-third. Counter-intuitively, even though Cap has improved, average attendance would drop. This is because going from one to two teams increases the number of home games by 100% while the total attendance has only increased by around 33%. In Brisbane’s case, if (big if) these numbers held, average attendance across the two clubs would be approximately 25,000, still well above the NRL average.

The great thing is that we can use this logic to (very) loosely examine expansion and contraction opportunities, the nerdiest of sports nerds pursuits.