Tuesday 6 October 2015

Voronoi footballs

Via Keir Clarke's Maps Mania I saw the latest voronoi designed map this morning. Made by Guus Hoekman, the map uses voronoi polygons to subdivide the world such that each polygon represents a spatial partition that contains one football club.

It's a neat coded solution for the novice football supporter to simply locate themselves and then determine which is their closest team. 

Why get all cartonerdy about it then? Well, partly because I've not had a good cartonerd rant for a few months (despite several maps causing much ire) but moreso, because maps like these never seem to go beyond providing a small technical solution to an overly simplistic question. The real world is far more complex than maps like this ever try to take on. Closest is not necessarily the way you choose your team.

Football (probably most sport fandom) is a hugely complex soup. There are three rules for supporting an English football team. First, you support the team you were born and raised closest to so in that respect the map could be deemed a reasonable effort. Proximity gives ample opportunity for the young supporter to shift allegiance if parents move around a bit as well, before your fandom settles. The point here is that you don't need a map to tell you you're local birth-right football team but being close doesn't always stack up either. County or city boundaries, rivers and other features can all modify proximity.

Second, you support the team your father supported. Before someone screams 'what about your mother's team? Well, to be honest your father should have already got that covered if he chose his partner wisely. This is always a good option if your parents moved to somewhere bereft of a decent team yet hailed from a footballing mecca. If neither of these solutions appeal, you simply support Manchester United because that's what you do if you've never been anywhere near Manchester.

Regardless of these basic rules, supporting a team is so much more than a geometric solution that a map can provide. It's about territory, local rivalry, physical and social geographies and all sorts of other real world dimensions that cannot be adequately represented by voronoi polygons.

The map has omissions. It fails to show the lower leagues of most countries. It only considers the men's game and for some countries, such as the U.S. the women's game is arguably more popular anyway.

The map allows the classic 'neutral' fan to select a team based on tier. Why would anyone bother with lower leagues? Let's take a look at how this plays out by selecting just the first tier teams in an area I am familiar with.

So in England, being from Nottingham, the map would encourage me to support Leicester City. An odd choice since local rivalry dictates this is impossible. Worse, the map suggests that I would likely be sat alongside someone from Derby. This is utterly absurd. And look more closely...the English Premier League teams are on the same map as the Welsh Football League. They just don't compare.

My team is Nottingham Forest. I was born in Nottingham and raised 2 miles from the ground, on the right side of the River Trent to be a Forest fan. If my parents had decided to live on the other side of the river I'd have been a Notts County fan (shudder!)...or else I might have invoked the second principal and supported Chelsea since my father's lineage gave me that option. My mother's background was irrelevant...Rotherham United - footballing wilderness. I was fortunate when growing up that Forest happened to be one of THE most successful teams of the era (late 70s and early 80s in particular). What a fantastic quirk of location...I'd nailed it by luck alone.

Tragically, they've not been so good recently (by recent, I mean 20+ years and at least once I vaguely considered invoking the Chelsea lineage) and are now in tier two. According to the map below I am at least I'm properly deliniated from that lot from Derby and Leicester with perfect walls erected...though not quite in the right spot based on the geography of the places themselves.


Furthermore, most of Lincolnshire remains in Forest territory and this is wrong (unless you're my brother who sadly moved in this direction but at least maintains his footballing heritage). Adding tier 3 attributes the vast expanse of flat nothingness that is Lincolnshire to Peterborough United while Forest's polygon shrinks to make way.

But that doesn't stack up either. Peterborough's average attendance is 5,600 whereas Forest's is about 20,000 so the polygon doesn't really relate to the potential pool of support. Or does it...Nottingham is a much larger city in terms of population size (310,000) that Peterborough (116,000) though much of that expanse around the town is rural and sparsely populated. So, again, voronoi representations really don't adequately represent population distribution, structure or density of the real places from which support for a team is formed.

Finally when we add all teams into the mix and look at England's best supported team again...they are represented by one of the tiniest polygons. Most of Manchester United's fans live outside of this polygon and it's likely many of Manchester City's fans live inside it.

So voronoi's look nice, they are easy to make and when you have a point-based dataset you can compute them to demarcate space. Whether they make any sense whatsoever is down to understanding the input data and the questions you want the resulting map to support. In this case, the geography of football fandom is so much more complex than a voronoi can ever hope to show.

Want a more considered view of football fandom and how it is spatially formed? Check out James Cheshire and Oliver Uberti's Football Tribes map in their superb book The Information Capital. Football Tribes:

It's based on tweets (a dataset I've often been critical of) but heck...at least it demonstrates the complex structure of football fandom and how it is in no way possible to use a voronoi polygon as a way of reflecting on that geography.

Ultimately, data isn't just data. It has context. It often requires a deeper understanding and domain knowledge before you begin to represent or map it. Often, it's incapable of being used on its own to support meaning.

1 comment:

  1. Maybe it does offer at least an indication of local grown support, e.g. why someone from Warwick might be a Coventry City season ticket holder or why the population of Banbury might be comprised of an intriguing mix of Coventry, Northampton Town, Oxford United and Swindon Town fans? -- although I guess this itself shows the weakness of Voronoi as surely MK Dons is the nearest club!

    Anyway, I would like to ask, do you think it even possible to map football allegiances? The splat of twitter data above shows how the number of permutations that impact on someone's club and/or current location is almost infinite. Would love to know if you have a clever idea around this.