PageRank in March Madness
Each year in march, 64 of the best college basketball teams in America come together and compete in the NCAA Division I Men’s basketball tournament in hopes of claiming the title as the best team in the nation. Teams must make it through 5 rounds without a loss just to obtain the opportunity to play in the national championship game. With neutral court sites, loud fans, and one loss spelling elimination, upsets almost always occur throughout the tournament, giving the tournament its name “March Madness.” However, before teams can even have the chance to prove themselves in this tournament, they have to be awarded one of the 64 available slots.
The NCAA is divided into 32 conferences. The winners from each conference is automatically awarded a spot in the tournament. The remaining 32 spots are decided by the NCAA selection committee after taking into account a team’s performance, wins, losses, point margins, road record, strength of schedule, player availability, and even popularity amongst fans. Each year, the committee is often subject to lots of scrutiny as fans do not agree with their selection of teams. Fans disapprove of the fact that members of the committee get to decide which teams are more deserving than others. With the current selection methods being more subjective, many believe that a more objective process is needed. The PageRank method can be a solution to this issue as it gives an accurate ranking of teams that is based on past and predictive performances.
In order to implement this PageRank method, a data set was created that included winning team, losing team, point margin, date of game, and venue. A network of teams was then created linking the losing team to the winning team. A red arrow link represented a loss and a green arrow link represented a win. Additionally, the edges were weighted so that the value of each game could be more accurately represented. Once the graph was completed and all the links were formed, the PageRank algorithm was utilized to rank the teams. Teams were assigned PageRank values taking into account both the number and the quality of their links. With this new method, the first 32 slots were saved for teams that won their conference and the remaining non-conferences winers were selected depending on their PageRank scores.
Source:
https://content.iospress.com/articles/journal-of-sports-analytics/jsa200425