HOW DOES IT ALL WORK? provides fan sentiment analysis on English and International Football.

We trawl through thousands of tweets every day to extract the overall mood associated with teams and players, using state of the art artificial intelligence algorithms.

  • Tweets
  • Tweets

  • Teams

  • Players

  • Hashtags

What is a Sentiment score ?

The Sentiment score is a trusted measurement of public approval of teams and players, taken from the tweets of millions of English Premier League followers.

The Sentiment score represents the ratio between positive and negative tweets over a certain period of time (we use 3 days). So if six fans tweet positively about a football player and four tweet negatively about him, the Sentiment Score is 60% ( 6 / 6 + 4).

A Sentiment Score is calculated after a team or player has received at least 100 associated tweets with either positive or negative sentiment.

We use a smiley-scale for the sentiment score

smiley extremely happyWhen the Sentiment Score is above 75 %: excellent sentiment

smiley happyWhen the Sentiment Score is between 60% and 75%: good sentiment

smiley confusedWhen the Sentiment Score is between 50% and 60%: mixed sentiment

smiley unhappyWhen the Sentiment Score is between 25% and 50%: bad sentiment

smiley extremely unhappyWhen the Sentiment Score is less than 25%: very bad sentiment


We have created a database containing the official Twitter accounts and hashtags associated with all twenty Premier League teams and players.

  • Fan tweets are retrieved in real-time through Twitter’s Streaming API, and then filtered by hashtag for each of the teams.

  • For the sentiment analysis, we exclude tweets that contain another team’s hashtag in order to eliminate statistical noise contained in them.

  • The full list with more than 400 hashtags and other insightful twitter statistics can we found here.
TeamHashtag 1Hashtag 2
Man City#comeoncitizens#comeonmancity
Man Utd#GGMU#ManchesterUnited


Results from academic research show evidence that Twitter contains enough information to be useful for predicting outcomes in the football games.

With Twitter’s 330m users (src: Statista) creating 500m tweets per day (src: Internetlivestats), social media is a data rich resource that has great potential for sentiment analysis, so we can better understand the mood on a given subject. We built a prediction model that takes into account both Twitter-derived Sentiment Score and the betting odds. Predictions are based on this model. If the sentiment-based Win probability is higher than the Win probability implied by odds, we bet on the selection.