|
|
|
|
SportsQuant Ratings ExplainedThe SportsQuant rating algorithm was originally developed in 2004 to rank college football teams, and in 2006, the method was extended to college basketball. Ranking a large number of teams using results from a small number of games is difficult even when the objective is obvious. However, the ranking question is inherently ill-defined. We want to determine the best team(s). But what is the criterion? Many ranking methods are based solely on wins and losses. Others are based on point-scoring and margin of victory. Which are better? Both paradigms have their shortcomings. Win-loss models often predict that undefeated teams (even those who played a very weak schedule) will never lose and are infinitely better than even the best one-loss teams. At the other extreme, point-scoring models ignore winning entirely and, consequently, can rank a mediocre team with a few blowouts ahead of a solid, yet unspectacular team with a better record against a tougher schedule. Our rankings are the first to combine both wins and losses AND point-scoring data to extract the most from a limited and unbalanced schedule of games. In addition to considering win-loss and point-scoring data, our rankings explicitly account for home-field advantage and implicitly consider strength of schedule. Rather than rehash the details here, we refer the curious to Annis and Craig (2005) for in-depth explanation of the ranking procedure.
Our rankings were mentioned in the November 18, 2005 Kansas City Star. Until recently you could read about them at the newspaper's website, but News isn't news if it's old, and the article has been relegated to the archives.
Ratings vs. RankingsThe SportsQuant rankings are based on ratings for each team which are estimated using our algorithm. Ratings assign a numeric value to each team's offensive, defensive and overall strength. Rankings, on the other hand, are just an ordering of teams from best to worst (in this case, based on teams' overall ratings). Rankings tell us that one team is better than another, while ratings give some indication of how much better. A ranks can be computed easily from the set of ratings, but ratings can not be computed by knowing the ranks. An advantage of the SportsQuant ratings is that they provide a means of predicting future outcomes (e.g. What is the probability that Florida beats Georgia?, What is the expected number of points that Purdue will score against Indiana, etc.).
Schedule StrengthWe recently introduced the SportsQuant strength of schedule ratings. Check them out. |
Send mail to
webmaster@sportsquant.com with
questions or comments about this web site.
|