Many fans will rely on hunches and luck when predicting winners in the NCAA men’s basketball tournament. But students and faculty in NDSU’s statistics department began crunching data to help pick winners when the field was announced on Sunday.
Gang Shen, assistant professor of statistics, and Rhonda Magel, department chair, worked with students using a variety of statistical techniques to complete brackets for the tournament.
Students working on the project include Yingfei Mu and Bryan Rask. They expanded work done by former students Xiao Zhang and Sam Unruh.
The students already have been compiling data on teams, such as NDSU’s men’s team, who have earned automatic berths in the tournament. Using seasonal averages in statistical categories such as assists, turnovers and defensive rebounds, the group was able to calculate the chances of winning between every team in the tournament.
Results will be updated throughout the tournament, which ends April 7, to show how well the predictions stand up. “The bracketing is based on estimated probabilities from models and is not going to be completely accurate since the probabilities of 63 games needed to be estimated,” Magel said.
One technique involves a conditional logistic regression, Magel said. A similar approach was used for last year’s tournament, and students and faculty have fine-tuned their approach and included the women’s bracket this year.
Magel and student Wenting Wang developed three equations for the women’s bracket for predicting the point spread in games based on seasonal averages of various game statistics.
Magel said they’re aiming for approximately 70 percent accuracy in predicting the game’s outcomes. The brackets are displayed on bulletin boards on the second floor of Morrill Hall.
No. 12-seeded NDSU is scheduled to play No. 5 seed Oklahoma in the first round of the tournament at 6:30 p.m. on Thursday in Spokane, Wash.
NDSU is recognized as one of the nation's top 108 public and private universities by the Carnegie Commission on Higher Education.