As football season approaches and before anyone places his or her bets they may want to consult Saif Ahmad ’12 and Michael Jannings ’13 to see who will win. The two students spent the summer at the College of Wooster figuring out a way to identify the most dependable variables of the football outcome and create mathematical models to provide reliable predictions about the NFL season outcomes.
The two students worked with John David, visiting assistant professor of mathematics and computer science and Drew Pasteur, assistant professor of mathematics and computer science this summer during the Applied Mathematics and Research Experience (AMRE) program this summer on campus.
AMRE was created by the College’s department of Mathematics and Computer science to help students gain experience with the practical applications of mathematics outside† of the classroom, in ways a traditional classroom setting could not provide. The students, advisors and sometimes clients or businesses involved work together for eight weeks to answer their team’s research question.
The neural networks system is based solely on numbers and according to David, “the most objective way to predict NFL games.” Pasteur and David came up with 11 statistical comparisons per team, in addition to a home-field advantage; for a total of 23 variables. David said, “What really matters is how the two teams match up…Say one team has a great passing offense and its opponent has a weak secondary, while the second team has a good offensive line, and the first team has a week defensive line. What happens? How will that affect the outcome? Our model is designed to learn to predict that.”
Ahmad and Jannings were suggested their topic of using neural networks to predict NFL outcomes by the instructors and began working right away.† The beginning was a challenge, however. Ahmad is from Jamaica and not very familiar with the NFL and, according to Jannings, “the hardest part of our research was learning the background material needed to work with neural networks.† Neither Saif nor I had any experience with neural networks so we had to at least get a basic understanding before we could start working.”
However, once the two students became familiar with the method they both became very interested in it.† Jannings stated, “One of the most rewarding parts of our research came when we realized that our work had the potential to compete with other experts in the field who had a much better understanding of the sport than we did. It’s kind of neat to be able to demonstrate the power of math and computer science tools by themselves.
More and more major companies are using the neural networks approach to predict outcomes.† “What inspired us to pursue this are the underlying networks and how they function,” said David.
The research team consulted ESPN.com for statistics of the 2007 and 2008 football seasons in order to see if they could predict the outcomes of the 2009 NFL season. After choosing the statistics wisely, Ahmad stated, “We were right with the experts…our predictions had been correct around two thirds of the time in terms of accuracy, which is comparable to the experts at ESPN.com.”
“Well, it was a very good experience for me and Saif,” said Jannings, “I think we were both excited to be able to apply our knowledge and skills towards something that a lot of people would appreciate or at least find interesting.”† The team hopes to introduce new variables into the research in order to become even more accurate in their predictions.