Twitter predicts SPOTY winner Wiggins

bradley wiggins wins BBC sports personality of the yearSunday nights Sports Personality of the Year awards were one of several television events this year which Twitter users probably couldn’t avoid; with over 322,000 tweets mentioning the awards ceremony (either by name or the #SPOTY hashtag) on the day of the awards, 280,000 during the broadcast itself, peaking at just over 6,500 tweets per minute as Bradley Wiggins was announced as this years winner.

And, as with voting events ranging from the US elections to the X-Factor final (and our own look at the Mercury Music Prize), people are looking at how data from social media can be used to create audience insights. A reasonable hypothesis might be that the members of the public who were interested enough to cast a vote would be well represented by the Twitter users interested enough to tweet about their favourite candidates.

To look into this, we collected tweets mentioning “Sports Personality of the Year” (or the #SPOTY) hashtag and looked into what they were saying.

On the night of the event itself, we saw a familiar pattern – a large spike of mentions as the broadcast began, with additional peaks around points of key interest (ie. award winners), and a build up of additional Twitter volumes at 9:30pm, after all of the nominees had been shown to the audience and the public vote began.

An initial analysis might be to look at the relative volumes of mentions of each of the twelve award nominees during the period that voting was open. This shows a fairly reasonable picture of where the actual vote went;

Wiggins – the winner of the public vote – saw more mentions than any of the other nominees, and Ennis came in second in the ranking. But Murray (3rd place in the awards) saw fewer Twitter mentions than Mo Farah – so not a perfect prediction.

Why might this be? Well, one hypothesis might be that SPOTY voters (and Murray’s supporters in particular) are not well represented by social media users, so were less likely to be tweeting along with their votes. Or perhaps Murray’s fans were more likely to express their support on a different platform like Facebook (a platform which we have very little visibility of, due to the semi-private nature of the network.)

Another view to take might be to look back to the 26th November when the shortlist was announced, when the overall count of mentions of each of the nominees on Twitter was slightly different. Although it could have been used to accurately predict the winner (or even the top 3 places – although in the wrong order), Murray does notably better in 2nd place.

So what happened to the public perception between the announcement of nominees and the actual vote?

Well, perhaps nothing. There are a number of different reasons someone might tweet about something. One explanation might be that on the day of the announcement, the tweets were mainly about the news (so mentions correlated to who people thought were the most newsworthy nominees — ie. the most famous), but on the day of the awards the tweets were more about people talking about their own support. So it might be that Murray mentions on 26th November were simply from those who thought it was noteworthy, rather than supportive.

A third — and more accurate — picture comes from looking at Twitter volumes before the voting lines opened. During the build up to the show, although volumes were lower, Wiggins still clearly dominated mentions. But during the broadcast itself, three athletes very clearly stood out as being the most mentioned – the three winners. Although less clear from the chart below (which shows higher intensity of tweeting clearer than the total volume), the volume of mentions here not only stood as an accurate prediction of the three winners, but also of the order in which they ranked.

Interestingly, Twitter reactions over the course of the day turned out to be a better predictor than during the actual voting itself.

But, whichever way you look at it, Wiggins remained ahead throughout. But this particular look reveals a clear reaction to the videos shown about the three eventual winners that was notably greater than the other nominees. Conveniently, the format of the broadcast (with a short video of each athlete explaining why they were shortlisted) makes it easy to see how the reactions to each athlete differed.

However, this somewhat simplistic analysis actually overlooks the true level of support – while there were 13,918 mentions of ‘Wiggins’ during the voting period, an additional 9,048 tweets mentioned ‘Wiggo’, with just a few hundred mentioning both. This reveals an additional complexity in using social media analysis as a form of quantitative measurement – without being able to set up a level playing field, it can be difficult to accurately compare two different sets of data.

Scott Thompson, digital research manager, Starcom MediaVest Group