Mapping the Barassi Line

Revision as of 06:43, 22 July 2022 by Tenniscourtisland (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Wikidata Fellow Brett Tweedie on his project exploring the great football divide in Australia
, Brett Tweedie.

In 2022, Wikimedia Australia offered three $1000 (AUD) Wikidata Fellowships to curate a data set, develop a prototype or undertake an investigation using Wikidata supported by an experienced mentor. Wikidata Fellow Brett Tweedie is a data visualisation designer and developer (who played exactly 4 games of rugby in high school and was terrible in all of them). Supported by mentor Sam Wilson and Toby Hudson, Brett describes his project exploring the great football divide in Australia.

Explore the full project website here.

“Australia is divided by a deep cultural rift between the north and the south, known as the Barassi line. It runs between Canberra, Broken Hill, Birdsville, and Manangrita [sic], and it divides Australia between Rugby and Rules”

Ian Turner, 1978[1]

I'd first come across the notion of the "Barassi line" about a decade or so ago, while working on a project about the relative popularity of Australia's four football codes, and was reminded of it again on a road trip my wife and I took towards the end of last year. Over the course of a few weeks we drove from Sydney out to Broken Hill and back, with a few little detours over the Murray along the way. As we drove, I noticed the goalposts on the ovals we'd pass gradually change from the H-shaped, two-posts-joined-by-a-crossbar of rugby, to the four posts of Aussie rules. And because I'm a nerd and easily sidetracked, it got me thinking about how hard it would be to try and work out where the Barassi line currently was, or even if there was such a clean dividing line between the two codes. Anyway, thanks to Wikimedia Australia for letting me have a crack at working out roughly what the Barassi line looks like today.

The idea was to try and plot the line using the locations of all the Aussie rules and Rugby league clubs in Australia. Turner's original conceit compared Aussie rules to rugby in general, so both league AND union, but to keep it more manageable, I just focused on Aussie rules and league. The data was gathered from a variety of sources such as club and competition websites, but also from Wikidata, Wikipedia, news articles, sport results sites, community websites, fan websites, local councils, OpenStreetMap, Google Maps, social media etc.

In the early stages I did a quick test using the home ground data in OpenStreetMap just to get a rough idea of how many data points would be involved, as well as the rough distribution (but didn't use this data for the final maps, as I couldn't be sure that it covered all clubs). Each of these clubs were then assigned to one of the locations in the latest Suburbs and Localities on the ABS website, and from this I then calculated the centroids of each location. At the time of writing the maps included 2,380 football clubs - 1,519 Aussie rules and 862 Rugby league - in 1,865 different locations (as some had multiple clubs). Based on the number of Aussie rules vs Rugby league clubs, locations were then marked as either ‘Aussie rules’ or ‘Rugby league’ depending on which code had the greater number of clubs (or ‘equal’ if the number of clubs were the same). Then, in order to move from a discrete number of points to polygons covering the whole of Australia, I used a Voronoi diagram. This creates cells which denote the area where a particular point, in this case a place where a particular code of football has more clubs, is the closest, and therefore (in theory at least) the dominant code in that area.

Early concepts for The Barassi Line project

Obviously there's different approaches that I could have used, and although I tried to ensure the data is accurate, with so many different sources, there's still likely to be a few mistakes, such as clubs I've missed, clubs placed in the wrong location etc. Also, all clubs have been treated equally, regardless of size, and the centroids are obviously an oversimplification of the actual towns and suburbs, plus the Voronoi cells generated are based purely on geographical distance, rather than actual proximity by travel time. However, despite all this, the sample size is large enough that the overall picture is still (hopefully) fairly accurate.

For a while there I had grander plans of getting the dates the clubs were established to see how the line moved over time, but it took so much longer than I thought just to get and clean the initial data that I dropped that bit (but might have another go at it later on). There was also a fair bit of experimentation in the code along the way about the best method to generate the finals line(s), such as factoring in the dominant sport in neighbouring cells, or taking the average of clubs within a given (perhaps adjustable) radius etc, but in the end I just kept it fairly simple.

Testing the Voronoi diagram. Red areas (based on the colour of the Sherrin football, the official ball of the AFL) are where Aussie rules was dominant, blue (colour of the state where it was first played in Australia) is for league.

For those interested in the nuts and bolts of it all, it was built using Turf.js for the Voronoi diagram, generating the centroids, grouping the polygons, plotting the final paths etc, Leaflet and Tangram for the maps, and D3 for the rest. The data (kept in a rather messy spreadsheet) was then added to Wikidata using OpenRefine.

Thanks again to Wikimedia Australia for the fellowship, and thanks heaps to Caddie, Sam, & Toby for all their help along the way!

Note: It's been pointed out that even the term 'Aussie rules', tends to be used only by those north of the line (where I'm from), but have used it here given that was the term Turner used in his speech.



  1. "The greatest game (Barassi Memorial Lecture, Prahran College of Advanced Education, 1978)." Overland, (76/77), pp. 32–41
Discuss this page