Ozzie Sport

This is the last of my data dumps from this weekend.  I’ve heard that the penetration of the Alexa toolbar in Australia is really low.  Given that, these numbers are probably less than reliable.  Still, given the absence of other public data, you work with what you have.  This is one of those cases.  Because of some issues, this only includes NRL and AFL urls.  (My actual list is a bit longer and includes more teams.)

Related Posts:

• More popular Australian (and Kiwi) sport related Facebook fanpages and groups (November 23)
• More popular Australian (and Kiwi) sport related Facebook fanpages and groups
• Most popular Australian athletes, clubs, leagues and sport organizations on Twitter (October 11)
• Most popular Australian athletes, clubs, leagues and sport organizations on Twitter (October 4)

My dissertation topic involves doing a demographic and geographic study of Australian sport fandom online. There are several sites and social networks where you can get publicly available demographic data to begin to formulate a picture of the user population, and then segment that population out by interest in a league, sport and athlete. I’ve spent a lot of time looking at Twitter, Facebook and LiveJournal. Recently, partly because of a trip to the Wikimedia Foundation and discussions with a few people at UCNISS, my interest in who was contributing to Australian sport wiki articles on Wikipedia increased.

Finding out who edited Wikipedia articles using publicly available information is a bit of a challenge. The most reliable information for who edited comes from IP address information. IP addresses can provide an idea as to the geographic location of the contributor. It is easy enough, with the help of a friend, to create a tool that pull the history of a Wikipedia article, get a list of IP addresses that edited the article, feed the IP address into another tool that will pull up the general location of the contributor. (One of my favorite visualizations of this type of information is WikipediaVision.) The data isn’t always accurate and if I was looking primarily at New Zealand, a country without its own dedicated IP address range, this would be even less reliable. Still, for my purposes, this data works pretty well.

This data is still pretty limited. There are a lot of articles that are edited by non-anonymous users. Sometimes, it is possible to get demographic and geographic information about Wikipedia contributors by viewing their profile pages. This can just be time consuming to do manually if an article has a large number of contributors as you need to view a lot of user pages. It becomes a deterrence for trying to collect geographic information about article contributors.

I was looking for a more time effective and accurate method of collecting geographic and demographic information about contributors that is publicly available on their user pages. The easiest and quickest way to get this information on a mass scale is to utilize user box information. Many user boxes, when included on a user page, put the user into a category. These categories are often then linked through the Wikipedian category structure. Beyond that, user boxes involve templates. It is easy to get a list of articles (user pages) that the template is included on.

The methodology that I selected from this point is rather straightforward. It involved:

1. Select a category.
2. Copy and paste the list of articles (user pages) in the selected category to an Excel spreadsheet. Sort the list alphabetically. Copy and paste only the user pages to Notepad. Replace * User with blank. Copy and paste this list back to Excel.
3. Create a filter where the cell contains / . Select those cells. Copy them to notepad, replace / with [tab] in order to remove user subpages from the list. Copy this back to Excel. Select only the column with usernames.
4. Run an advance filter in order to remove all duplicate rows.
5. Copy this list back to the dedicated spreadsheet. Label all those users with the category from which they were pulled in a unique column.
6. Repeat steps 1 to 5 until all the categories that you want to have included are included.
7. Merge/Group all the rows by username.

This method may not be the most efficient way of going about doing this. It can probably be improved by automating some of these steps. In my case, step 7 was not able to be completed using Excel. I had to e-mail the file to @woganmay, who I believe converted the file to a mySQL database, used the group feature, converted the results back to csv and e-mailed the file to me.

In my case, I did not complete this for every category. Some categories did not seem worth it time wise as they had too few user pages to be included. In other cases, the categories were just too big to do. This included all the members of User de, User en, User es, User fr, User it, User jp. Only a selected number of categories were included because of time constraints. Data gathering was focused on categories that I perceived would have the greatest number of Australians and other possible contributors to Australian related articles. When these categories were more exhausted, categories with between 1,00 and 5,000 articles were selected.

There are all sort of limitations to this data. First, not everyone includes userboxes on their profile pages. This means that there could be a lot more Australians on Wikipedia than indicated by userbox inclusion on a user page. The assumption for the resulting data is that proportional representation exists for various categories. So while there are X amount of Christians and Y amount of Atheists, the assumption that the relationship between X and Y will always be proportional to the actual population on Wikipedia. Whatever data is available thus has to be viewed as good enough or supplemented by going to individual user ages to see if other information is available when a user appears where no information for someone when running against the history of the article.

Second, even when they do exist, there are often useful pieces of information that are missing. For example, in an Australian context, there is a userbox for Rugby League fans. There is not however a userbox for Australian rules footy fans. There are also not user boxes and categories for fans of NRL or AFL teams. (This type of user box and category exists for National Hockey League teams.)

About halfway through this process, I realized that this data could be useful for analysis beyond who is editing Wikipedia. At the moment, I’ve only totaled data I have for Australians. It is pretty fascinating and would be neat to go further with: How does the proportional size of the Australian Wikipedian population compare against the actual population? Does the size of the Australian Atheist versus Christiah community actively reflect the proportions in Australian society? Or is the Australian Wikipedian community demographically distinct from the greater population?

The following tables include the data based on people who were included in Wikipedians in Australia and its subcategories and Australian Wikipedians. A copy of the raw data can be found at October 9 – Wikipedia English Data – Australians.xls. The data is provided without comment though any attempts at explaining the patterns found are very much appreciated.

Related Posts:

• Online Activity in the Wake of the Melbourne Storm Controversy
• Melbourne Storm and Wikipedia
• Most popular Australian (and New Zealand) sport related Twitter accounts (November 24)
• More popular Australian (and Kiwi) sport related Facebook fanpages and groups (November 23)

I’m pretty much done with Twitter.  By that I mean that I’ve compiled my list of Twitter accounts.  I’ve got a user geographic list.  Neither one is likely to be improved.  All I need to do now is run those accounts through my program, then count up where people are from, map that and write my dissertation chapter.  I intended to do that soon.  After I finish my Literature Review.

In the mean time, I’m playing with Wikipedia.  I’m interested to know who exactly is editing Australian sport articles.  I can do that some what easily with anon-IP address edits.  When it comes to logged in users, that’s a bit harder.  I need to be able to get information off contributor user pages to determine that.  I don’t know who edited what.  To make my life simpler, what I did was go category like Wikipedians by location, and put that information into an Excel spread sheet.  Various categories were combined.  (I don’t have all categories of Wikipedians by…  I have a random selection of larger categories.)

I’ve got some technical issues in combining rows on Excel.  (I don’t know how to do it.  I’m waiting for a friend to get on MSN so I can harass them to do that for me.)  Still, in the process of doing this…  I figured I would post what I had found so far.

I’ve identified 2,220 people who claim to live in Australia.  The tables below give an idea as to the current count of Australians that fall into these categories.  If you don’t see a category, don’t assume it means there are zero people in it.  It could mean that or it could mean I haven’t processed that row.  It could also mean that I didn’t include that category yet.  It could mean that a category for it doesn’t exist on Wikipedia.  (Example: I have yet to see a category for Aussie rules football fans.)

Another thing to bear in mind, this data is from userboxes.  A contributor has to put a userbox on their user page.  Not everyone does that.  Some people may put userboxes up for some categories like language but not others like alma mater or religion.

That said, findings so far:

So if you’re an Aussie Wikipedian, you likely went to ANU, have a BS, are a current university student, are an Atheist, of English descent, work as an engineer, speak Greek in addition to English, are a Rugby league fan, are a Wikipedia administrator and rollbacker and use IRC.

Need this complete to see how well this stands up.

Related Posts:

• Expanded Profile of Australian En-WP users
• The Impact of Jason Akermanis’s Comments on the Western Bulldogs’s Online Fanbase
• Traffic to W-League Related Wikipedia Articles by Date
• An interview with the CEO of the Canberra United

Canberra has traditionally been a place that people aren’t from.  Rather, it is a place that people have traditionally moved to.  When they move, they take their sporting loyalties with them.  For that reason, there is a conflicting mess of sport loyalties in the city.  If you went to see the Sydney Swans play the Western Bulldogs at Manuka Oval this year, you could clearly see these different loyalties on display as it pertains to the AFL.  On the whole, I’ve generally found Canberrans to be pretty good with these loyalties: You can safely wear your team’s kit with out fear that some stranger is going to take the piss out of you.

The Grand Final is soon approaching.  Given the different sporting loyalties, it is interesting to see which team people in Canberra are barracking for.  Is one team more popular than another?  Do Canberran loyalties mirror national loyalties?

One way of measuring loyalty and team identification is to count how many people follow a team on Twitter.  All followers combined, Collingwood 7,381 followers, St. Kilda has 4,492 followers, Geelong has 2,153 and the Western Bulldogs have 1,554.  When Twitter followers for each team are filtered by location using time zones and user listed location, you can get an idea as to how many Canberrans are followers of the various teams playing in the Grand Finals:

It isn’t that big of a surprise that Collingwood is number one. Roy Morgan Research had them as the most popular Melbourne based team in 2009.  Beyond that, Canberrans on Twitter buck the popularity trend.  Roy Morgan has Geelong at number 2.  On Twitter, Canberrans are the least likely to follow Geelong.  Roy Morgan had the Western Bulldogs last.   Canberrans had them at third.  Overall, when compared to total number of followers for a team, Canberrans don’t  always follow the national patterns.  Again, Geelong and the Western Bulldogs are the ones that don’t match: More people follow Geelong on Twitter but more Canberrans on Twitter follow the Western Bulldogs.  It looks like Canberrans, who don’t have their own AFL team, follow a different drummer than the one that beats on a national level.

Another way to measure team loyalty and team identification in Canberra is to use Facebook.  Facebook says that there are 198,500 users who live in the ACT and 175,900 people who live within 50 miles of Canberra.  What Grand Finals teams do Canberrans like?

Well, not the Western Bulldogs, that’s for sure.  Canberran fans on Facebook much more clearly follow the national patterns as established by Roy Morgan.

Other social networks exist and Canberran sport fans are present on them.  One such network is LiveJournal (and its clones).  There are a number of fans on it who list their teams as an interest on their profile:   4 for Collingwood,  22 for Geelong,  16 for St. Kilda and 23  for the Western Bulldogs.  When broken down by city and state, there are two fans from the ACT who list Geelong as an interest and one fan who lists the Saints as an interest.  This pattern for the ACT does not mirror the pattern for the whole of LiveJournal.  It also doesn’t follow the pattern for team fandom size as identified by Roy Morgan Research.

Over on 43 things, there is one Canberran who has a goal related to an AFL team.  It just isn’t a team in the finals.  (The Canberran has a goal of buying a Sydney Swans jumper.)  On bebo, there are 18 people identified as being from the ACT who list an AFL team as an interest, four of which list a Grand Finals team as an interest.  Two support Collingwood, one supports Geelong and one supports St. Kilda.  This pattern mirrors the national one as established by Roy Morgan Research.  On Blogger, there are four people from the ACT who list an AFL team as an interest.  Of these four, only one lists a finals bound team: Geelong.

What does tell us?  Canberran AFL fans are most likely to be found on Facebook.  Their support of their clubs is similar to the national club support though there are differences.    Different social networks attract Canberrans fans of different teams.  It shouldn’t be that hard to find a like minded fan who will support your team if you’re here for the game.  The easiest time of doing that will probably be for Collingwood Supporters but Bulldog fans shouldn’t be that worried either.  Just go on Twitter Doggies fans and reach out to your fellow Canberrans.

Most of the raw data referenced in this post can be found at csv.ozziesport.com/.

Related Posts:

• Based on data to date: Biggest Australian sports fandom
• Most popular Australian (and New Zealand) sport related Twitter accounts (November 24)
• More popular Australian (and Kiwi) sport related Facebook fanpages and groups (November 23)
• More popular Australian (and Kiwi) sport related Facebook fanpages and groups

I feel guilty as I’m kind of posting as I do with out any particular pattern. Anyway, I looked at Google.com.au and friendster (both of which have a load of garbage unrelated results) and recorded the totals for number of results for each team. I also did that bebo, care2 and blogger. (I also did ecademy and blackplanet but results were 0 the whole way through.) I’ll see if I feel like exploring and recording the size of the Perth sport fandom some more. The following table includes the results so far. If you have methodology questions, let me know.

Related Posts:

• Data dump: Australian sport on bebo
• 43things.com and Perth based sport clubs
• Most popular Australian (and New Zealand) sport related Twitter accounts (November 24)
• Most popular Australian (and New Zealand) sport related Twitter accounts (November 18)

I haven’t done a real data dump day in a while and today appears to be the day. I want a distraction and trying to figure out the most popular sport teams in the Perth area fits the bill. The first is a list of goals for each time on 43things.com.

Related Posts:

• Data dump: Australian sport on bebo
• Most popular Australian (and New Zealand) sport related Twitter accounts (November 24)
• Most popular Australian (and New Zealand) sport related Twitter accounts (November 18)
• Most popular Australian (and New Zealand) sport related Twitter accounts (November 16)

I’ve been some what lax with collecting data during the past month.  This morning, I decided to remedy this a bit by updating my Alexa rank list for Australian sport clubs and teams.  This is what can be found below.  (The list isn’t all comprehensive.  I tend to mostly focus on the AFL and NRL.) It is interesting to see who is doing well, especially when compared to other clubs in the same league or clubs playing in the same geographic area.

Related Posts:

• Most popular Australian athletes, clubs, leagues and sport organizations on Twitter (October 11)
• Most popular Australian athletes, clubs, leagues and sport organizations on Twitter (October 4)
• Most popular Australian athletes, clubs, leagues and sport organizations on Twitter (version 2)
• Most popular Australian (and New Zealand) sport related Twitter accounts (November 16)

Yay! It is over. This is the final total. Rankings changed through out the tournament. For the two sites, there were audiences I’d like to call distinct. The data when examined in total suggests that different fan communities inside a wider community of football fans use geolocation services differently. As previously mentioned, this may be useful for people involved with the Rugby World Cup in 2011. I should also see about getting some of my AFL and NRL data out… but that’s not as on point.

The winner of the World Cup was not the winner for average checkins per game they played. English speaking countries didn’t dominate at the top of the ladder. You didn’t need to be a team who made it through more than the first round to crack the leaderboard’s top 10, though it certainly helped. The home team didn’t necessarily hold a major advantage.

Foursquare

Gowalla

Related Posts:

• World Cup, through quarter finals: Foursquare and gowalla checkin totals
• World Cup, end of the second round: Foursquare and gowalla checkin totals
• World Cup, end of the first round: Foursquare and gowalla checkin totals
• World Cup, day 8: Foursquare and gowalla checkin totals

The World Cup is done.  Based on averages, there do appear to be trends for usage around different teams.  I don’t have any other large international competition which to compare this two but it might be of interest to others who are organizing such events like the women’s World Cup for soccer and the Rugby World Cup.  Some national contingents will be the same and if we assume that these patterns hold across nationalities… it might give national sides, companies and organizers using geolocation services an idea which one is better for their audience and which games to target.  This is especially true if we assume that Foursquare and Gowalla users are representative of active social media users period who will use other services to discuss the event.

Related to the results: For the Germany vs. Spain game,between getting data before the event and after the event, Foursquare merged two locations.  If I had just used the one, it would have possibly included an additional 43 checks for the game that were from previous games.  Thus, I subtracted the previous venue’s totals from the venue that they were merged into.

Onwards with the results…

Foursquare

Gowalla

Related Posts:

• World Cup, end of the second round: Foursquare and gowalla checkin totals
• World Cup, end of the first round: Foursquare and gowalla checkin totals
• World Cup, day 8: Foursquare and gowalla checkin totals
• World Cup, day 7: Gowalla checkin totals

That Ghana versus Uruguay: Wow. Wow.  Also, wow!  It was a major win for Foursquare as the game smashed the record for total checkins set by South Africa for the first game with 676 checkins compared to 335.  The game booted South Africa and Mexico from the top of the Foursquare table.  Sadly, it didn’t do this so much for Brazil and the Netherlands.   Gowalla sadly didn’t see the same Foursquare response.  There just doesn’t seem to be the same audience and there isn’t the same level of deviation of for national sides and between rounds.3.8 average checkin for the top team on Gowalla and 0 for the team on the bottom.  Foursquare has 179 on top and 17.3 on bottom.

There appears to be something going on with Foursquare.  I’ve included at the bottom a table which demonstrates what I’m talking about.  It looks like for four venues Foursquare removed a number of checkins or a large number of people deleted their own checkins.  I’m not certain what exactly happened as I can’t see who checked in and what games were effected.  It is annoying.  (And I don’t know who else is covering Foursquare and the World Cup to see if they’ve noticed these issues.)

Foursquare

Gowalla

Foursquare Checkin Losses