Data analytics has already predicted the winner of Super Bowl XLVIII even before kick-off.
A Microsoft researcher renowned for accurately predicting major events has analysed all the relevant data to predict the most likely outcome of this year’s Super Bowl final at MetLife Stadium, New Jersey on February 2.
David Rothschild, an economist at Microsoft Research in New York City, has carefully examined the two consistent data sources for predicting the outcome of sporting events, to discover that Denver Broncos are on course for a win – maybe.
He said: "First, I’ve looked at prediction markets, where investors buy and sell shares of contracts over who will win the game. For example, there are contracts right now that are going to be worth $1 if Seattle wins the Super Bowl and $0 if Seattle loses the Super Bowl. The amount of money investors are willing to pay for that contract is highly correlated with the probability they will win the game.
"Secondly, statistical models based on wins, field, turnovers, etc have historically been very strong."
This year, Rotchschild says the data sources are pointing to a narrow win for the Broncos but it is going to be an extremely close call.
He explained: "This year both the markets and the statistical models are showing something remarkably similar: a real toss-up. My latest is the Broncos at 53% to the Seahawks 47%."
These numbers are likely to change as the we get nearer to game time, but Rothschild describes his previous predictions of major events as "extremely accurate and well calibrated."
He added: "I was very happy to see my political models correctly forecast 50 of 51 Electoral College outcomes in February of 2012. A much harder challenge, I was happy to see my models provide the correct answer in 19 of 24 categories in last year’s Oscars. And, I have been pleased to see my models have two perfect weekends in a row this year heading into the Super Bowl."
In order to accurately make such predictions, he must ensure that he is answering the question that most stakeholders want or need to know the answer to.
"In American football, that is the probability of either team winning the game. Second, I want to make sure that I have considered what my goals are for the prediction in terms of accuracy. Third, I ensure that I consider how the data will allow me to update the predictions continuously (i.e., in real-time through the event.)."
Finally, he puts the data through a rigorous course of statistics – econometrics and/or machine learning, depending on the question.
"Then I have a model that takes raw data and creates an accurate, relevant, and real-time updating prediction."
Keep an eye on PredictWise.com to see how Rothschild’s predictive numbers progress in real-time on Sunday.