Page Ranking Sports Teams
Lisa Zyga wrote the article, “New Algorithm Ranks Sports Teams like Google’s” to describe a new form of ranking sports teams, based on Google’s PageRank algorithm. The term PageRank itself is named not after the pages it ranks, but instead after Larry Page, one of the founders of Google. This system of ranking web pages on Google utilizes the link relationships between pages, thus resulting in ranks based on how many links direct towards their pages, and the rank of such pages that link to said page. This directly relates to our in class study of how Google ranks pages to show up as a result of entered searches.
In a paper written by Anjela Govan, she outlines the problems of ranking based on simply wins and losses as it is simplistic. This method is carried out by the Colley Matrix Method as it takes into account only wins and losses, yet is still frequently used. The Keener Ranking Method takes into account scores, with the use of a score smoothing component to prevent the manipulation of rank through running up their scores. Additionally, rankings based off of polls are laden with bias and personal intuition, which Feng believes has no place in determining the ranks of sports teams.
Ed Feng, a researcher in statistical mechanics at Sandia National Laboratory is developing a new algorithm for sports team rankings, called Power Rank. In his adaptation of Google’s algorithm, he uses only two aspects of Football game results: the game score and home field advantage. He argues that this prevents human bias, and memories from last season from influencing rank unfairly. Additionally, this would provide more accurate results as opposed to simply using the win-loss records of teams. The links between the nodes (football teams) represent games played between the teams, and each link is directional. The number assigned to the links are based on the score and location of the game (home field advantage is represented as three points in the NFL), with the number estimating the probability that the team at the head of the arrow will beat the team at the tail on a future neutral site. Determined by their season performance, each team is ranked based on their value, just as Google’s PageRank orders the pages after entered searches. In accordance with PageRank’s algorithm, Feng’s algorithm allows teams that beat highly ranked teams to gain more value, than if they were to beat a weaker team. While there may appear to be prediction potential, Feng emphasizes that it is to be used only for ranking past performances.
This article illustrates the wide scope of influence which Google’s PageRank can have in various arenas that require numerical assessments of its constituents, particularly in the sports world. This comprehensive algorithm takes out a significant amount of bias, while still being able to take relative strengths of teams into account when assessing scores in wins and losses.
Sources:
Article by Lisa Zynga: http://phys.org/news/2009-12-algorithm-sports-teams-google-pagerank.html
Paper by Anjela Govan: http://meyer.math.ncsu.edu/meyer/ps_files/sasgf08rankingpaper.pdf