Test your Super Bowl picks with this new machine-learning prediction tool

Put the power of random forest regression to work for you

WSO2 BigDataGame machine learning football super bowl

Machine learning can be your friend in this year's Super Bowl.

Credit: WSO2

There's been no end in sight to the advance of machine learning into the world of enterprise software, but this week a new online tool debuted for the purpose of sheer fun: predicting the winner of the Super Bowl.

Built on WSO2's open-source Machine Learner technology for predictive data analysis, BigDataGame uses Apache Spark and random forest regression to compare teams and make predictions.

It will be Carolina v. New England in the big game, the tool predicts, and Carolina will have the edge with a 50.82 percent chance of winning. New England's chance is 49.18 percent. In the meantime, anyone can put in their team picks for the playoffs and see what data science would predict for an outcome.

The company's blog says development of BigDataGame was motivated by a single question: "Given everything we can find on football teams and the Big Game, can we use our software to predict the winners?"

To create the tool, WSO2 tapped historical data from Pro-Football-Reference.com for the 2012, 2013 and 2014 seasons and tested a series of algorithms by trying to predict the results of the 2015 season.

So far, it boasts a 76.5 percent accuracy rate, and because it's constantly learning, that should keep improving.

"This program isn’t perfect," the company warns. "It’s very much a hobbyist project at WSO2, and we’re still in the process of tweaking it."

Factors such as injuries and morale are impossible to predict, but the math is sound, WSO2 says. After every game, it plans to update the system with the latest data.

It all serves ultimately as a showcase for Machine Learner, which launched in November and is available separately or as part of WSO2 Data Analytics Server 3.0.

Related:
Shop Tech Products at Amazon