Social media, particularly Twitter, has become the place to go to if you want to find out what’s happening right now and what people are talking about, but is it possible that it can do a little more than that? Can it tell us not just what’s happening, but what’s about to happen too?
Before Saturday’s Eurovision Song Contest took place, Minttwist had a go at analysing activity on YouTube, Spotify, Twitter, Instagram and Facebook in an attempt to work out who was going to win the competition. It’s a method that has seen the site correctly predict the 2013 and 2015 winners.
When all the numbers had been crunched, the predicted top ten was:
When the votes (eventually) came in for Saturday’s contest in Sweden, the actual top ten was:
On the whole, it wasn’t a bad guess; the analysis has correctly predicted two of the top three songs, albeit not in the correct order, and six of the top ten. There are anomalies though, such as winner Ukraine not featuring in the prediction at all, which was picked up on as unusual by the site as it was one of the favourites. Also, while the Netherlands and Malta only just missed out on the top ten, Italy was some way off in 16th and the Spanish entry fared poorly in 22nd.
The analysis shows that it is possible to make an educated prediction on the result of the competition, but it’s a somewhat mathematical one and there are, as the article points out, a number of caveats. The performance order can make a difference, and this study also doesn’t take into account that someone may give a particularly good or bad performance on the night.
In effect, social media and search analytics can be used to make predictions in what is a development of probability theory – if something is talked about often enough, it becomes more likely to happen, or at least it indicates that it’s the expected outcome.
Prior to the FIFA World Cup 2014 in Brazil, search engines Google and Bing went head-to-head in predicting the winners of the knockout ties, and Bing came out on top forecasting 15 out of 16 winners, while Google called 14 results correctly. These stats appear impressive, but bear in mind that the knockout stage games largely went as expected, and that both search engines wrongly thought Brazil would beat the Netherlands in the third-place playoff even though they had been battered 7-1 by Germany a few days earlier. Also, it’s worth noting that Bing’s success rate in predicting the group stage games of the same tournament was an unremarkable 60%.
It’s interesting to see the limited success that can be gleaned through studying online activity to predict the outcome of events taking place away from the internet, but at this stage, there are a lot of human variables that can’t be fully factored into the equation.