Twitter Bracketology: Using Social Analytics to Predict March Madness Upsets

    By Topsy | Small Business

    This year’s NCAA March Madness tournament officially kicks off on Thursday with the Round of 64. Here at Topsy, we always like to look to the Twitterverse to see if they possibly know something the rest of us don’t. Using our Twitter sentiment analysis we made predictions for each game based on overall sentiment for each team over the past 30 days. We took a weighted average of the sentiment, putting more emphasis on recent days, to come up with a single score for each team.

    Below is the bracket we submitted to our internal office pool. Were we smart with some of these big first round upsets that includes Davidson over Marquette and Northwestern State beating Florida? Not sure, but the folks taking to Twitter sure did seem to think so by (NW St got a sentiment score of 88 to Florida’s 74. Scores range from 0 to 100, with 50 being neutral, 0 being the most negative and 100 the most positive).

    Twitter Bracketology: Using Social Analytics to Predict March Madness Upsets image Screen Shot 2013 03 19 at 12.24.42 PM 648x1024Screen Shot 2013-03-19 at 12.24.42 PM

    Is powerhouse Duke really going to lose to lower ranked Creighton in the Round of 32 or even more surprising, number 1 seed Indiana falling to Temple? Our sentiment analysis thought so based on a score 83 for Duke vs. 90 for Creighton, and a score of 83 for Temple vs. 80 for Indiana.

    Yuck. Indiana weakness is toughness … and NC STate and Temple are tough teams. Won’t be easy, Hoosiers.

    — Jason McIntyre (@TheBigLead) March 17, 2013

    A quick sentiment analysis comparing the schools in our Final Four shows that it was a dead heat for most of the last few days with Louisville barely squeaking by Gonzaga over the past day.

    Twitter Bracketology: Using Social Analytics to Predict March Madness Upsets image Final Four Sentiment 1024x225Final Four Sentiment

    What makes predictive analytics work so well is the scale of the social conversation we’re able to capture. Each team receives thousands of Twitter mentions per day, and aggregating sentiment from all of the individual tweets tends to bring out the main themes. Although some of our bold predictions might seem crazy at this point in time, we have had great success in the past predicting the outcome of the presidential race by state as part of our Twitter Political Index and the Oscars where we correctly predicted 5 out of the 6 Oscar winners.

    Stay tuned as March Madness unfolds to see how close our predictions are and updates we make to the outcomes along the way. For now though, Let’s go Cards!

    More Social articles from Business 2 Community:

    Subscribe to our mailing list
    * indicates required
    Small Business Services