PageRank Applied to Basketball Tournaments
This article discusses PageRank as an alternative to current methods used to forecast performance for NCAA tournaments and to decide which teams play in the tournaments. The authors talk about how PageRank as an algorithmic system is a much more objective measure compared to the current methods that involve a council and can be swayed by factors such as a team’s popularity. They used a directed network to connect teams throughout the season with edges indicating that the teams played a game and the arrows pointing to the team that won. The PageRanks were created and gave weight to edges based on factors, specifically mentioning the weight of edges based on “point margin, date, and venue.” From these factors, they used point margin to not overly penalize teams who lose in blowout games. They also assigned a higher value to teams who perform better later in the season and closer to tournaments. They gave less value for winning home games versus away games as winning home games is more typical.
PageRank proved to be effective in predicting performance of teams, as tested with the Midwest region’s PageRank values and how they compared to their actual performance. This connects to ideas we learned in class about the value of the weight of PageRank in graphs. The contents of this article were relevant to what we discussed in class about using directed links and PageRank values to assign authority scores; the teams with greater authority were the ones expected to perform better in tournaments. This article assigned weight of PageRank in a more complex way, not just accounting for the total shares a node receives, but also weighing certain factors that better predict team performance in tournaments. This additionally proves the effectiveness of weighing edges in networks to solve real problems and predict outcomes. I think it would be interesting to see how the results here could apply to other sports teams or other scenarios where past performance can indicate later success.
https://content.iospress.com/articles/journal-of-sports-analytics/jsa200425