Friday, November 4, 2011

ESPN the Mag

With many thanks to John Gasaway, I am pleased to announce that my preseason rankings appear in ESPN the Magazine’s Nov. 14th College Basketball Preview. If you are not familiar with my preseason predictions model, here is an overview:

The best predictor of future success is opponent adjusted margin-of-victory. This is the basis for Jeff Sagarin’s Predictor Rankings and Ken Pomeroy’s Tempo Free Rankings. But college rosters experience significant turnover between seasons. Fortunately, Dean Oliver developed statistics that estimate each player’s contribution to the offense and defense. These individual “tempo free” statistics can be used to estimate how each team’s offense and defense will change from one year to the next. Today I use this information to predict the 2011-12 season.

The basic model incorporates several well-established basketball facts: First, experience matters. Teams that have more returning minutes (and possessions) tend to improve. Second, teams that return more efficient scorers (i.e. better shooting percentage, fewer turnovers) improve more than teams that return less efficient scorers. Third, the biggest leap in development is from a player’s freshman year to his sophomore year. Teams that give major minutes to freshmen tend to improve significantly the following season. Fourth, the loss of injured players (such as USC’s Jio Fontan), and the return of injured players (such as Purdue’s Robbie Hummel) has a predictable impact on team performance. Fifth, for incoming transfers, the performance with the previous team provides some information about the player’s future performance. Sixth, coaching ability impacts performance in a predictable manner. And finally, high school recruits can have a significant impact on a team’s performance. High school recruits ranked in the Top 10 have the biggest impact, but players ranked in the Top 100 are also important. This year I also account for the fact that historical team prestige impacts recruiting. (This is less important for BCS teams where the biggest factor is Top 10 and Top 100 recruits. But for smaller schools, historical team prestige is often the only factor we have to separate the quality of recruiting classes.) These factors are combined to produce a numeric ranking of teams for the upcoming college basketball season.

ESPN voted for the Top 25 and only used my rankings to rank the other 319 teams. For the most part, I think ESPN got the top 25 right. But if you have questions about why my model loves or hates certain teams, send me a tweet @DanHanner.