Dukies, Single People, And Women: Data Show Who Sucks At Picking BracketsS

In a feel-good "enormous corporations share user-data" story, Facebook and Yahoo have joined forces to give us a rare look into the demographics that drive March Madness picks. Drawing on data from the 60,000 users who used Facebook to login to Yahoo's College Tournament Pick 'Em, Michael Bailey (creator of Facebook's March Madness Fandom Map) and Sean Taylor (creator of the NFL Fandom Map) showed what kind of people–by region, college affiliation, gender, relationship status, and level of education–have made terrible picks so far, and what groups made slightly less terrible picks.

You can read the whole article on Facebook, but here are a few highlights:

  • Yep, fandom matters: Users who liked a basketball program's Facebook page picked that team to win 1.4 more games in the tournament than average. Liking the school's athletic page (but not the basketball page) added 1.0 win, while being an alumnus added 0.6 wins.
  • The Midwest is kicking ass: The Midwest has had a strong tournament so far, with four teams (Louisville, Ohio St., Michigan, and Marquette) cracking the Elite Eight. As a result, users in Nebraska, Minnesota, Wisconsin, and Michigan are some of the most accurate in country. Users in Oregon have ridden their local program's "upsets" to considerable success, and Louisiana has been very accurate as well, because voodoo. The East, and especially North Carolina (stung by Duke's and UNC's losses), have done particularly poorly.
  • Single people and women filled out crazier brackets: Everything you learned in Single White Female is confirmed. These groups predicted (not necessarily correctly) more upsets, while college grads played it the safest.
  • Men picked better: On average, male users have correctly picked 1.5 more games than have female users. Conversely, every woman who is filling out a bracket for the first time ever is currently winning her office pool.

This is just a taste, graphs and charts abound in the full article.