The Gladness and Sadness of March Madness

Dan Grady's picture
 By | March 18, 2014
in WebFOCUS, Analytic Dashboards, analytics, Business Intelligence, customer survey analysis, dashboard, feedback anlaytics, march madness, sentiment analysis, Social Analytics
March 18, 2014

March is my favorite month.  The weather starts to warm up (well, at least it used to). We honor St. Patrick’s Day, which happens to double as my anniversary. My eldest daughter was born in March. And in addition to all these personal milestones and celebrations, March plays host to my favorite sporting event, --the NCAA Men’s Basketball Tournament, otherwise known as March Madness.  

Student athletes playing their hearts out in a survive-and-advance tournament, combined with the David and Goliath matchups that produce Cinderella stories every year, creates an atmosphere that can’t be beat.  It’s a galvanizing event that brings together friends, families, and coworkers to participate in office pools, where we all use our own strategies for filling out our brackets.  (As a side note, the ritual of the brackets have such a pull on us as a nation that a 2013 study estimated that the first two days of the tournament alone cost American companies $134 million in productivity dollars.)

 

Since, I’ve been focusing on Sentiment Analysis of social media, I thought it would be interesting and hopefully entertaining to turn my tools toward the public’s emotions around the brackets. WebFOCUS’s Sentiment Analysis capabilities did such a good job indicating the eventual winner of the Oscar for Best Picture, I wanted to take the same approach in determining the actual winner of this year’s tournament.  If you aren’t familiar with sentiment analytics, it’s one of the ways we help our customers drive insights out of the vast amounts of textual data they are collecting.  We can look at a set of text and based on the tone (positive or negative) of the text we give it a numeric score. Once it is quantified, you can use the value for additional analysis. For example, sentiment scores are particularly helpful in prioritizing responses to customer feedback.

We are calling this analysis “Gladness, Sadness and March Madness” because it’s all about finding emotion in the text. To do this, we calculated sentiment scores for all of the posts and comments on the Facebook page for each of the 68 teams participating in this year’s tournament, for one week leading up to the tournament (Mar 9-17).  Each team has a certain percentage of posts that scored positively on their page and a certain percentage of posts that scored negatively (and some posts that didn’t register with the engine).  For the Oscars prediction, the ultimate winner, “12 Years a Slave,” had the highest percentage of negative and positive posts (71%). We’re going to use that same rule to move the teams through our brackets here.

As expected, our sentiment scores created several upsets, which you can see in the brackets in the image gallery above.  Here are some highlights:

  • In the opening round, #16 seed and play-in victor Albany pulled a stunning upset over top seeded Florida – 53.25% of the posts on Albany’s page scored to only 25.56% on the Gators page
  • Many of the pundits have picked the Michigan State Spartans as a likely winner this year. They survived the first round of March Madness but were eventually defeated by Harvard  --  45.75% of Harvard’s posts against the Spartans 32.55%
  • There was a defensive struggle in round 1 between the Oregon Ducks and BYU, with BYU winning by a score of 25% to 13.24%
  • There were two buzzer beaters in the Elite 8 – Albany continued its Cinderella run by upsetting Stanford 53.25% to 52.25% and Harvard slipped past UW – Milwaukee 45.75% to 44.59%
  • We wound up with a Final Four that no one would have expected if they were going on pure basketball knowledge – Albany, Harvard, Kansas St. and Creighton.
  • The Creighton Blue Jays took on the Cinderella story from Albany and took home the title having a 60.05% sentiment score to Albany’s 53.25%.

I have to say, I went to the Big East tournament over the weekend and the Creighton fans were by far the most passionate at Madison Square Garden, so after seeing all the numbers I am not that surprised.

Some additional stats:

  • The University of New Mexico had the highest percentage of negative posts (10.8%)
  • Michigan had the highest number of post and comments for the week with 7,913
  • New Mexico State University didn’t have an overly active page with only 25 posts and none of them scoring negatively with the engine.

While I am not going to use the sentiment engine to assist in filling out my brackets -- and I don’t recommend you do either -- this exercise highlights a few key pieces of functionality available within the WebFOCUS BI and analytics platform:

  • The ability to use social data to do comparative or competitive analysis.
  • The ability to quantify textual data by applying sentiment scores to help get more value out of the data.  In this example, we applied the sentiment scoring engine to social data, but it can be applied to any textual data such as that found in feedback you are receiving from customers, citizens, patients, etc.

Sentiment is only one way we help our customers analyze feedback style data. We also offer word frequency analysis and search-based applications.  If you are interested in learning more, I am presenting a Daytime Demo on this topic on March 26.

That’ll do it for our 2014 March Gladness, Sadness and Madness Tournament.  I wish you all good luck in your office pools for the real tournament. I’m in several myself…I always go with Izzo and the Spartans, and I like them again this year, but I don’t like to pick with Dick Vitale and the rest of the experts, so I am going with Virginia in most of my brackets.