For football fans across the globe, the World Cup is the highlight of the football calendar. It is, however, a point in time where thousands of dreams are shattered as countries drop out of the competition – some rather spectacularly. But when your country has perhaps been eliminated, or maybe if you didn’t even qualify, where do your loyalties lie? Do you cheer for anyone else? Your geographical neighbour – or perhaps even who they are playing against? For instance, with the referendum looming, did the Scots cheer for England, or their opposition in the 2014 World Cup?
Well, Twitter can provide a few of the answers and the Text Analytics Group research laboratory at the University of Sussex agreed to team up with website The Conversation to try to find out.
As part of his PhD in Computational Linguistics, Simon Wibberley’s research involves developing software that can analyse social media posts to see what users think about a certain issue – a system known as Method51. Its full name is “methodologically sound sociological research on Twitter,” but the team cropped it back to the first six letters and the number of characters that made up the rest of the full name – in line with Twitter’s own character obsession.
For The Conversation project, the Sussex team set a couple of parameters. First, they only collected tweets during the game and one hour either side – and for this project, only those tweeted in English. Then Method51 was instructed to divide the tweets into “cheers” for either side and irrelevant posts by creating bespoke document classifiers for each game, whilst also splitting the tweets by geographical location.
The team had to disqualify a lot of tweets – throwing away all those that are either irrelevant or not geographic. For theEngland v Italy game on June 14 for example, they started with 616,000 tweets during the time frame that had been set. Once classified however, only around 50,000 “cheers” one way or the other remained.
The Uruguay game on June 19 got 511,000 tweets, but this time as many as 91,000 classified as positive to one or the other team. By the Costa Rica game on June 24, which made no difference to either team’s progress because England were out and the central Americans were through, the number of tweets fell to 140,000, of which fewer than 7,200 expressed a preference.
As you would expect, tweets coming from England were heavily in favour of the country – and increasingly so with each game. Against Italy, there were about four times the number of “cheers” for England as for Italy. In the Uruguay game, the “cheers” were six to one in favour, and against Costa Rica they were over ten to one.
With the caveats mentioned earlier about foreign language countries, the same was true in reverse with England’s opponents. In the Italy game, tweets coming from Italy were around 3.5 to one in favour of their country, in the Uruguay and Costa Rica games, the tweets coming from the two countries were respectively around three to one and 17 to one in favour of home.
Interestingly, on the occasions when these countries were not playing against England, their tweets swung behind the English on every occasion. With England the poorest performers throughout, maybe they all wanted England to do their national team a favour by defeating another rival.
It is worth pointing out that a large number of tweets came from Brazil, and were heavily weighted towards England each time, possibly by English fans actually at the tournament. The US, which was the other big tweeting nation, started mildly against England but was cheering them loudly by the third game.
When the team turned to the results from the other three UK nations and Ireland, the picture gets even more interesting. During the Italy game, only the Welsh wanted England to win, and much more lukewarmly than home fans – 1.5 to one in favour compared to the four to one in favour in England. Northern Ireland was about 1.1 to one in favour of Italy, while the Irish and the Scots were almost equally negative at 1.4 to one for the Italians. The Scots were a shade more indifferent, posting just under 2,700 tweets compared to more than 3,200 from Ireland (despite its smaller population).
In the next two games, the growing bias in favour of England in the English tweets also happened in the other home nations. Against Uruguay, England’s six to one bias was widest (bar Brazil, at nine to one), but Wales and Northern Ireland’s support both rose to almost 2.5 to one in favour. The Irish were now shouting almost two to one in favour of England, while even the Scots tilted to 1.3 in favour. With qualification now at stake, the whole of the British Isles got behind the English team – if only just.
Against Costa Rica, the sample sizes outside England became very low due to lack of tweets. But the results are eye-catching nonetheless: Northern Ireland out-cheered even England, 14 tweets to one in favour. Wales was more than four to one in favour and Ireland remained at two to one. As for the Scots, they threw caution to the wind and went a bit England daft, cheering five to one in favour of their southern neighbours.
The team’s conclusion? The Scots were certainly less keen on England victory than any other part of the British Isles. To this extent the old maxim is true, albeit it’s almost as true in the Republic of Ireland. Yet neither did the Scots want England to be eliminated by Uruguay, even if they cared less than the other home nations. And once this came to pass, they definitely didn’t want Britain’s only representatives to go home empty-handed…
(images and data courtesy of Simon Wibberley)