March Madness Is Here. Are You Ready To #HackTheBracket Again?

College basketball is big news every March, when more than 5.6 million viewers on average per game watch the “March Madness” NCAA basketball tournament.

Along with the amazing games the tournament produces, roughly 70 million people create brackets to predict the outcome of each one. Given all the guesswork involved, the odds of randomly filling out a perfect bracket are about one in nine quintillion! (Compare that to the odds of winning the $1 billion Powerball lottery last October, which was about 1 in 300 million.) Mathematicians have tried to use models to improve the odds, but even that doesn’t guarantee a perfect score.

Last year we decided to get involved in the fun. For the first time ever, we opened Adobe Analytics to the general public for free so fans could better predict March Madness winners based on which team had a statistical advantage in each category. And it worked: our bracket generated leveraging Adobe Analytics data finished in the 98th percentile.

This year, we’re bringing it back. And, it’s better.

The success rate of Adobe Analytics-generated brackets is no surprise to us, considering the plethora of data it can analyze, such as height, offensive rating, and free throw percentage. Adobe Analytics also slices and dices the data however you want it to.

Using data from Sportradar, we loaded into Adobe Analytics the intelligence of more than 56,000 games with over 100 key metrics that could be used to predict winners. We included just about everything — from three-point shooting, to effective field goal percentage, to offensive and defensive rebounding.

As a result, the way one fan looks at the data will be very different from what another fan looks at. Our hypothesis: when you uncover what is leading to the NCAA players’ and teams’ best performances with Adobe Analytics, your odds of creating a successful bracket skyrockets.

Here’s what’s different this year:

• Last year we had game summary data, and this year we are loading in play-by-play data for games where it is available. For this season, we will have somewhere around 3.5 million rows of data to play with, compared to the 50,000 or so rows last year. Similar to last year, there will be approximately 100 dimensions and 100 metrics that you can use in your analysis.
• With the play-by-play data, users will now have the ability to create shot charts.
• This year we bring in a new metric we’ve invented named “Watchability”, which looks at various ‘story’ aspects of game—how many big swings the game had, how close the game was, how well the teams played, if a player had a great individual game, etc—and assigns a score from 1 to 100 based on how much fun that game would have been to watch.
• We also brought in location data for all of the games this year, so users can visualize where a team has played games.
• There are new metrics for those who want to take a less serious approach to picking their bracket – this includes arena capacity, undergrad enrollment, among others.
• The experience in the Analysis Workspace itself should be a lot easier to use this year. A new “Dropdowns” feature added to Adobe Analytics last year will allow more simple interactions within the dashboard, whereas last year was more focused on doing deep ad-hoc analysis. Read this tutorial to take advantage of the dropdown feature.
• The play-by-play data also allows users to look at who performs best in “clutch time”, exclude “garbage time” and look at new ways to analyze game summaries.

Basketball fans everywhere: we invite you again to #HackTheBracket, and let’s see if we can bring our bracket success rate even higher. Click here to get started. And see how you can win tickets and travel to this year’s championship game in Minneapolis, Minnesota on April 8.