Students Create Olympic Crystal Ball
November 11 2009
Remember sitting in class, wondering what the heck the teacher was talking about and how would it apply to your life? Often school can be a rigorous and hard to apply to real life, however, Westminster College freshman Claire Bonner and Karl Gerner created a statistical program for their final management project that turned out to not be another lame assignment.
Originally thinking they could help their college ski team out, they developed a statistical model that was so unique that it has helped the U.S. Snowboard team make decisions on where specific halfpipe athletes should compete, and which athletes should ultimately make the 2010 Olympic team. The USSA recognized:
In the past, U.S. Snowboarding’s athletes attended as many World Cup events as possible in order to ensure athlete qualification, as there was no data available to show which competitions were most advantageous to attend and which were not.
“We decided to create a model to help U.S. Snowboarding get rid of those educated guesses,” Bonner explained.
While the U.S. Ski Team and U.S. Snowboarding has been a partner of Westminster College for nearly five years, U.S. Snowboarding first approached Westminster College in 2008 to ask for help with forecasting its athletes’ competition schedules to maximize the U.S. team’s Olympic chances. With the team’s domination of the 2002 and 2006 Olympic Games, it hoped to continue a strong performance while balancing its athletes’ busy schedules. A key part of that is ensuring that U.S. athletes have strong results on the World Cup circuit which is used to determine national quotas for each of the three snowboarding Olympic events. The goal was to maximize the quota spots in as many events as possible.
Using a spreadsheet-based software suite for predictive modeling, forecasting and simulation called Crystal Ball, the students put together a program that could determine which athletes should compete and where they needed to compete to get maximum results while minimizing costs and interference to the athletes’ professional competition schedules. Since many snowboarders juggle their schedules with major events like World Cups, U.S. Snowboarding Grand Prix, the U.S. Open and X-Games, it was essential to develop a system that could help them balance their season.
After working tireless hours in the library and seeking guidance from their professor, Dr. Alysse Morton, the students completed what they considered to be a “perfect” model for the team. The model is capable of taking the athletes’ previous, present and future predicted scores, as well as the scores from other competing athletes from around the world to predict whether the athletes would qualify for the Olympics or not.
“The model works by predicting a range that each athlete could score in a given competition based on their previous results in similar events, such as the World Cup,” described Bonner. “Then this score is multiplied by either a one, if they attended the event, or a zero if they did not attend the event. Then all of the athletes are put in order based on their ranking in the raw scores. The rank received is assigned with corresponding points to the athlete.”
While the program sounds relatively easy to understand, the duo had a few glitches along the way.
“The main problem we faced when working on this project was the actual gathering of the previous athlete data due to many of the athletes’ short history competing at the World Cup level. Also, snowboarding halfpipe is a relatively new sport in the Olympic Games, which made it hard to gather good data. We had to predict the athletes’ average performance and calculate the possible range that they could achieve at their maximum and their minimum potential,” said Gerner.
U.S. Snowboarding Program Director Jeremy Forster believes the model was very useful in planning the upcoming competition season.
“The work that was done was a great help in planning our season and the athletes’ schedules,” he said. “The initial forecasting was very useful and looks like it will be quite accurate once the final Olympic quota spots are determined on January 18, 2010.”
When asked if any of the results surprised him, he noted, “In general, I think the results were expected but the model helped validate our initial thoughts.”
All in all, the feedback the students received was very positive.
“Overall, I think they did a great job. Both Bonner and Gerner put a lot of time into understanding the goal of the project and researching the results of the athletes to make sure the data was as accurate as possible,” said Forster.
“All the feedback we have received from USSA has been extremely positive,” echoed Bonner. “I believe that they are truly stoked about being able to reduce their costs and have the ability to predict more accurately which athletes are going to qualify for the Olympics.”
Although it is still early to tell if this system is fool proof, Bonner and Gerner will continue to refine their numbers and make adjustments. When the final Olympic team is selected in January, the students will be able to tell just how well their model really worked.