At the launch of ESPN’s Basketball Power Index (BPI) in 2014-15, the No. 1 team in the country was clear. The Kentucky Wildcats were 11-0 and only beginning to hit their stride. At this point of the 2015-16 season, the race for the top team is more competitive. There are about 10 teams -- including Michigan State (11-0), Kansas (8-1), Iowa State (9-0), Oklahoma (7-0) and Purdue (11-0) -- that can make a claim to be the top team in the country.
Now that the college basketball season is more than a quarter of the way complete, ESPN’s Basketball Power Index (BPI) has enough information to settle these debates. It currently rates Purdue as the top team in the nation after it won its first 11 games, including winning three games against BPI Top 30 opponents by double digits.
Before digging into the Boilermakers’ impressive start, let’s get a further understanding of BPI and its features. BPI is ESPN’s college basketball rating system first launched in 2012 as a way to measure the overall performance of all 351 Division I men’s basketball teams (rated on a 0-100 scale). It can be used on both the game level and season level to answer questions such as which teams should make the NCAA Tournament and how each of those teams would perform in the big dance.
Like many other ratings systems, BPI accounts for outcomes of games, location (home/road/neutral) and quality of opponents, but BPI has at least one feature that differs from other system out there. Let’s look at each distinguishing feature through the lens of the 2015-16 season.
SOS of opponents -- Beyond W-L
RPI is the most notable college basketball ratings system and has aided the NCAA Tournament’s selection committee for years. One drawback of RPI is that it measures the strength of schedule (SOS) purely by opponent W-L. As of Tuesday, Incarnate Word has played RPI’s toughest schedule, and with three wins -- all against non-Division I teams -- the Cardinals currently rank 108th in RPI.
BPI (and most other systems) dig deeper into the true strength of each opponent. With eight games against Division I programs, including five games against teams ranked in the top 100 of BPI, UCLA has played BPI’s toughest schedule to date. Incarnate Word’s SOS is ranked 203rd by BPI because of its three games against non-Division I programs.
Scoring margin -- with diminishing returns for blowouts
A win is a win, but how a team wins can tell a lot about that team’s strength. Scoring margin is generally a better predictor of future performance than a team’s W-L record alone, so like most systems other than RPI, BPI accounts for how a team won the game.
BPI, however, decreases the value of a blowout: a 30-point win is rated about 20 percent better than a 15-point win, not twice as good. Four of the top five teams in scoring margin against Division I programs -- Louisville, Purdue, Oklahoma and Michigan State -- rank in the top 5 of BPI, with Purdue being the only team in the nation to win all of its games by double digits.
Notably absent from the top 20 of BPI are Kansas (24th) and Maryland (26th). Both teams rank lower than 15th in scoring margin against Division I opponents and have a BPI SOS ranked outside the top 100.
Pace of game matters
BPI takes score differential one step further by accounting for pace of play. Obviously a slow-paced team such as Virginia will not have as many possessions and not be able to match the scoring margin of an equally efficient, faster-paced team.
Net efficiency (offensive efficiency minus defensive efficiency) in each game is captured before adjusting for other factors such as game site, scoring margin, opponent strength and missing key players. The top four teams in BPI rank in the top six in net efficiency.
De-weighting games when teams are missing key players
Marcus Paige missed North Carolina’s first six games, including its 71-67 loss to Northern Iowa. When the Tar Heels’ résumé is presented to the NCAA selection committee in March, should those games be evaluated in the same way as they would be if he was healthy?
BPI de-weights games in which key players don't play. It accounts for key players based on minutes played and adjusts the importance of games when those players are missing and when they are not. For example, once Paige accumulates enough minutes to be considered a “key player” those games will be weighted slightly less for the Tar Heels and their opponents.
This de-weighting helps with prediction accuracy going forward. Although BPI was not designed to be predictive, it has performed as well if not better than other systems in predicting NCAA Tournament games.