Football fan locality- An analysis of football fans tweet locations Harris N 1, James P 1 1 Newcastle University, School of Civil Engineering and Geoscience, Newcastle-upon-Tyne, NE1 7RU November 11, 2014 Summary This paper looks at the validity of using social media as a dataset for spatial analysis and demonstrates the use of geo-located tweets to investigate the locality of football club fan-bases KEYWORDS: Social media, Locality analysis, Football 1. Introduction Twitter is a free social networking and micro-blogging service that enables its millions of users to send and receive messages of up to 140 characters (tweets). Twitter to date has over 284 million users and processes overs 500 million tweets a day. Many recent events have been documented live via Twitter users on the ground. For example the hashtag #OccupyCentral became a live feed on information of the recent protest in Hong Kong with 700+ tweets a minute being sent at the peak of the protests (Boehler, 2014). Approximately 3% of tweets also include location information in the metadata of the tweet. This amounts to 15 million tweets a day that contain location (Dredze et al., 2013). Twitter has been used to map the effects of an earthquake (Yin et al., 2012), the spread of disease (Lan et al., 2012) and many other applications. Tweets using the official club hashtags from football clubs provide a potentially rich data source to test the assumptions often made about the location of supporters. This is a hotly debated topic on forums and in pubs and assumptions about fan bases such as Manchester United supporters being predominantly based in the South East and their rivals Manchester City being locally based are widely quoted in the press (MEN, 2007). It is also a common assumption that the locality of football fans is inversely proportional to success, with teams that win trophies attracting fans from across the UK. (Dudley, 2014). 1.1 Football Tweets Data is collected by listening on the official hashtag of each club in the football league e.g. Manchester United's official hashtag is #mufc. The location, the team in question and the time, are captured and stored in a PostGIS relational database. To date we have recorded several million football tweets spanning the last 3 seasons. 1.2 Locality In order to understand the locality of football fans a definition of local is needed. In the food industry the term local is often used and definitions exist for the term local. CAMRA (2014) defines a local ale as one produced within 30 miles of the pub. Farmer s Markets in Clark County (2014) only count produce within 3 hours travel time as local. Waitrose (2014) considers local recipes as recipes using ingredients from the same county. 2. Analysis In order to analyse the tweets a subset of the data was created using only tweets made during the October, 2014 international break (6/10/14 00:00-16/10/14 00:00). This would discount tweets from fans travelling to and from away football matches and reduce tweets about a contentious incidents in a game like a dubious red card. A total of 3000 tweets were collected for the Premier League and 1500 for the Championship.
2.1 Tweet distance as a Euclidean measure Initially the distance between tweet and club was analysed, with the home ground of each club acting as their reference point. Figure 1 shows the results for Premiership clubs. The results of this analysis are inconclusive although they suggest that there is some truth in the assumptions with the bottom of the table filled with the glory clubs of Manchester United, Arsenal and Liverpool. More unexpectedly, it also includes Crystal Palace and Swansea City. Swansea s low score is probably due to their generic official hashtag of #swans. At the top of the table are some of the teams that have been recently promoted. Manchester City sits third perhaps suggesting that there is some truth in them being locally supported. Average Tweet Distance per Club Average Tweet Distance (km) 0.00 20.00 40.00 60.00 80.00 100.00 120.00 140.00 160.00 180.00 200.00 1st Leicester City 40.36 km 2nd WBA 50.97 km 3rd Manchester City 55.04 km 4th Southampton 62.62 km 5th Everton 66.73 km 6th Tottenham Hotspur 68.10 km 7th Sunderland 76.89 km 8th Stoke City 86.24 km 9th Hull City 89.20 km 10th QPR 91.17 km 11th Aston Villa 97.91 km 12th Newcastle United 98.72 km 13th Chelsea 106.36 km 14th West Ham 116.07 km 15th Arsenal 118.13 km 16th Crystal Palace 119.39 km 17th Manchester United 130.45 km 18th Swansea City 156.41 km 19th Liverpool 161.91 km Figure 1: Average Tweet distance per club Another assumption often quoted by football pundits (Football club and fan relationships better in lower leagues, 2014) is that Premiership teams attract supporters outside the immediate geographic area whereas teams lower down the division are more likely to be locally supported. To test this hypothesis the Championship tweets from the same weekend were analysed in the same fashion and compared to the Premiership clubs. Figure 2 goes some way to supporting this assumption with only one Premier League side making it into the top 10 and only 7 in the top half of the table. The worst performing club is surprisingly Blackpool FC which may be due to the fact that at the time of data collection Blackpool were heavily in the news as a result of public disagreements between the owner, manager and fans (SkySports, 2014).
1st Cardiff City 15.64km 2nd Wigan Athletic 18.34km Average Tweet Distance per Club Average tweet distance (km) 0 50 100 150 200 3rd Middlebrough 37.23km 4th Leicester City 40.36km 5th Reading 41.70km 6th Nottingham Forest 44.20km 7th Millwall 44.22km 8th Charlton Athletic 45.11km 9th Derby County 46.96km 10th AFC Bournemouth 48.66km 11th WBA 50.97km 12th Bolton Wanderers 51.57km 13th Brighton & Hove Albion 54.95km 14th Manchester City 55.04km 15th Southampton 62.62km 16th Rotherham United 62.76km 17th Everton 66.73km 18th Leeds United 67.50km 19th Tottenham Hotspur 68.10km 20th Sunderland 76.89km 21st Brentford 78.15km 22nd Norwich City 78.50km 23rd Fulham 83.04km 24th Birmingham City 85.77km 25th Stoke City 86.24km 26th Hull City 89.20km 27th QPR 91.17km 28th Huddersfield Town 94.58km 29th Ipswich Town 96.62km 30th Wolves 97.06km 31st Aston Villa 97.91km 32nd Newcastle United 98.72km 33rd Watford 103.39km 34th Chelsea 106.36km 35th West Ham 116.07km 36th Arsenal 118.13km 37th Crystal Palace 119.39km 38th Blackburn Rovers 125.15km 39th Manchester United 130.45km Figure 2 Average tweet distance per club Championship side Premiership s ide 40th Swansea City 156.41km 41st Liverpool 161.91km 42nd Blackpool 177.91km
To further test this hypothesis the average distance from tweet to club per division was computed and the results shown in Figure 3. Here the Championship clearly out scores the Premier League rivals with a difference in distance of 40km in average between the two divisions. Average Tweet Distance per Division Average Tweet Distance (km) 0 20 40 60 80 100 120 Championship 68.27km Premiership 108.39km Figure 3 Average tweet distance per division 2.2 Proportion of local tweets This basic analysis applies a 30 mile buffer around the club as a measure of locality, however if a club is more isolated, like Norwich City, it is possible to be more than 30 miles from the club but it still be the nearest football league club. Therefore a tweet was considered local if it was within 30 miles of the ground or was the nearest football league club. Figure 4 shows the results of this test. Again the results are inconclusive with Liverpool and Arsenal and Manchester United in the lower half but accompanied by some other teams like West Ham and Crystal Palace. At the top of the table we have Manchester City and a smattering of some of the smaller clubs.
1st Manchester City 76.02% 2nd Leicester City 74.51% 3rd Sunderland 70.65% 4th QPR 67.79% 5th Newcastle United 65.20% 6th Everton 63.04% 7th Hull City 56.76% 8th Tottenham Hotspur 56.14% 9th Chelsea 53.85% 10th WBA 50.00% 11th Aston Villa 46.00% 12th Southampton 45.40% 13th Stoke City 44.74% 14th Manchester United 43.52% 15th Arsenal 42.77% 16th West Ham 40.00% 17th Crystal Palace 36.14% 18th Swansea City 32.43% 19th Liverpool 23.70% Proportion of Local Tweets per club Proportion of local tweets 0.00% 25.00% 50.00% 75.00% 100.00% Figure 4 Proportion of local tweets Again this test was repeated including Championship sides with the results shown in Figure 5. This supports the hypothesis that teams in lower divisions attract a local fan base with only two Premier League clubs now in the top 10 and only 6 in the top half of the table.
Proportion of local tweets per club Proportion of local tweets 0.00% 25.00% 50.00% 75.00% 100.00% 41st Liverpool 23.70% 42nd Blackpool 16.00% 28th WBA 50.00% 30th Watford 46.95% 31st Aston Villa 46.00% 32nd Southampton 45.40% 33rd Stoke City 44.74% 34th Manchester United 43.52% 35th Arsenal 42.77% 36th Wolves 42.22% 37th Birmingham City 41.03% 38th West Ham 40.00% 39th Crystal Palace 36.14% 40th Swansea City 32.43% 20th Huddersfield Town 62.07% 22nd Norwich City 58.93% 23rd Hull City 56.76% 24th AFC Bournemouth 56.25% 25th Tottenham Hotspur 56.14% 26th Chelsea 53.85% 27th Blackburn Rovers 52.17% 29th Ipswich Town 48.00% 16th QPR 67.79% 18th Everton 63.04% 19th Rotherham United 62.79% Figure 5 Proportion of local tweets 4th Charlton Athletic 79.49% 5th Bolton Wanderers 77.96% 6th Reading 77.78% 7th Derby County 76.79% 8th Manchester City 76.02% 9th Leicester City 74.51% 10th Millwall 74.42% 11th Brentford 71.43% 12th Sunderland 70.65% 13th Leeds United 69.75% 14th Fulham 68.97% 15th Nottingham Forest 68.75% 17th Newcastle United 65.20% 3rd Middlebrough 84.38% 21st Brighton & Hove Albion 61.54% 1st Cardiff City 95.24% 2nd Wigan Athletic 87.50%
Finally the proportion of local tweets per division was looked at and the results show in Figure 6. Again the Championship scored significantly higher than the Premier League. Proportion of Local Tweets per Division Proportion of local tweets 0.00% 20.00% 40.00% 60.00% 80.00% 100.00% Championship, 65.22% Premiership, 49.11% Figure 6 Proportion of local tweets per divion 3. Conclusions This analysis of the tweets for this one weekend broadly supports the hypothesis that Championship teams have a more local fan base than their Premier League rivals. When the results from both the distance and proportion tests are combined (Table 1) the results suggest that there is some truth to the assumption that the more successful teams have a wider fan base. The top end of the table is dominated by lesser teams with Manchester City and Everton being the exceptions, both of which are a city s second team which may result in their high placing. At the foot of the table, whilst it is Manchester United which has the reputation for the most widespread fan base, it is actually Liverpool who performed worst in both tests. The two notable anomalies in the bottom five are Swansea City and Crystal Palace. It is unclear from the tweet data itself why Crystal Palace does so poorly. Pos Club Table 1: Summary of results per club Rank on Average Distance Rank on Proportion of local tweets Lowest rank score Average Rank 1 Leicester City 1 2 1 1.5 2 Manchester City 3 1 1 2 3 Sunderland 7 3 3 5 4 Everton 5 6 5 5.5 5 WBA 2 10 2 6 6 QPR 10 4 4 7 7 Tottenham Hotspur 6 8 6 7 8 Southampton 4 12 4 8 9 Hull City 9 7 7 8 10 Newcastle United 12 5 5 8.5 11 Stoke City 8 13 8 10.5 12 Chelsea 13 9 9 11 13 Aston Villa 11 11 11 11 14 West Ham 14 16 14 15
15 Arsenal 15 15 15 15 16 Manchester United 17 14 14 15.5 17 Crystal Palace 16 17 16 16.5 18 Swansea City 18 18 18 18 19 Liverpool 19 19 19 19 Given that the Twitter data is a simple point data set any number of further analyses could be applied to it depending on the nature of the selected concept of local. Looking at the proportion of supporters within a region e.g. the South East or considering network travel time rather than distance may also yield some interesting results. Lovelace et al. (2014) discuss the limitations of using social media as a data source in spatial analysis and whilst it is not without its uncertainties it is important not to dismiss it outright as a concept. The sheer volume of tweets provides a rich data source for the research community. Further information on football tweets can be found at: http://ceg-sense.ncl.ac.uk/footballtweet.
4. Biographies Neil Harris is a researcher in Geomatics at Newcastle University. Whose research centres around the collection, management, analysis and visualisation of geospatial sensor and social media data. Phil James is a Senior Lecturer in GIS. His research centres on the use of Geospatial data to solve and support Engineering problems. Phil is the academic lead on Newcastle s Urban Observatory that integrates real time urban sensors across multiple scales with geospatial data and engineering models using the Cloud.
5. References Boehler, P. (2014) How Hong Kong's #OccupyCentral became a global topic on Twitter. Available at: http://www.scmp.com/news/hong-kong/article/1604409/infographic-how-hong-kongsoccupycentral-became-global-topic-twitter (Accessed: 3/11/2014). CAMRA (2014) CAMRA LocAle. Available at: http://www.camra.org.uk/locale (Accessed: 6/11/2014). Dredze, M., Paul, M.J., Bergsma, S. and Tran, H. (2013) 'Carmen: A Twitter Geolocation System with Applications to Public Health'. Dudley, B. (2014) 'Sing when you re losing', Supporters Not Customers. Available at: http://supportersnotcustomers.com/2014/04/22/sing-when-youre-losing/. Farmer s Markets in Clark County (2014). Available at: http://ext100.wsu.edu/clark/agriculture/small-farms/farmersmarkets/ (Accessed: 6/11/2014). Football club and fan relationships better in lower leagues (2014). Available at: http://www.staffs.ac.uk/news/football-club-and-fan-relationships-better-in-lower-leagues-staffs-uniresearcher-suggests-tcm4233521.jsp (Accessed: 6/11/2014). Lan, R., Lieberman, M.D. and Samet, H. (2012) 'The picture of health: map-based, collaborative spatio-temporal disease tracking'. pp. 27-35. Lovelace, R., Malleson, N., Harland, K. and Birkin, M. (2014) 'Can Social media data be useful in spatial modelling?', GISRUK Glasgow. MEN (2007) Ben salutes Blue Manchester. Available at: http://www.manchestereveningnews.co.uk/sport/football/football-news/ben-salutes-bluemanchester-1117827 (Accessed: 6/11/2014). SkySports (2014) Sky Bet Championship: Blackpool chairman Karl Oyston insists protests won't work (Accessed: 6/11/2014). Waitrose (2014) The Waitrose Small Producers' Charter. Available at: http://www.waitrose.com/content/waitrose/en/home/inspiration/about_waitrose/the_waitrose_way/s mall_producers_charter.html.html (Accessed: 6/11/2014). Yin, J., Lampert, A., Cameron, M., Robinson, B. and Power, R. (2012) 'Using Social Media to Enhance Emergency Situation Awareness', IEEE Intelligent Systems.