Ever wondered how the Media Buzz scoring system works? Below we take a look under the hood to see how each article scores are generated.
Media Buzz (MB) is media coverage which earns the player points based on headlines that the player generates. In theory at least, the more positive the headline, the greater the Buzz Score.
So how do FootballIndex assess what articles are positive and which are not?
To do this the site uses of an existing sentiment analysis model to analyse and then score the words contained within each headline. The analysis model (the opensource AFINN-165) it should be noted is not specifically intended to be used for football or indeed the sporting world.
It originated as a way to read sentiment from posts made by users on social media.
AFINN-165 model contains a word list of 3382 words rated from the very negative (-5 score) to the very positive (+5 score).
For example ‘breathtaking’ scores the maximum of 5 points where as ‘catastrophic’ scores a low -4 (the lowest -5 scores reserved for swear words and worse).
Each word in a headline is referenced against the list and generates a raw sentiment score accordingly for each word that matches. Words contained within a headline not found in the AFINN model are ignored. See here for a full list of the 3382 words used in the model.
Finally, a base score of +20 is awarded for each headline created regardless of the sentiment contained within it.
The formula then for Buzz Score is
(sentiment score +1) * 20
So using the example headline
‘Perfect’ generates a sentiment score +3 and ‘Improve’ a score of +2. No other words generate negative or positive scores.
This is then calculated as: (3(perfect)+2(improve)+1) * 20 (base article score) = a total Buzz Score of 120