New ESPN BPI Rankings are Useful but Far From GroundbreakingPosted by EJacoby on February 13th, 2012
Evan Jacoby is a regular contributor to RTC. You can find him @evanjacoby on Twitter.
The Worldwide Leader is again looking to stake its claim in the advanced stat revolution, this time in the college basketball realm. Saturday was the unveiling of ESPN’s new College Basketball Power Index (BPI), which ranks all Division I teams 1-344 based on a number of factors that go beyond wins and losses. The two most obvious questions to ask of this new system are: How does the BPI compare to the KenPom and Sagarin ratings that college basketball purists have come to know so well? And is this BPI ranking system any good on its own? These rankings appears to be quite similar to those of the popular KenPom, though there are a couple of unique additions to this system that attempt to make it stand out.
It’s hard to argue with what ESPN is doing here by releasing a brand new metric at the perfect time now that college basketball begins to own much of the sports spotlight for the next month and a half. It will be helpful to read ESPN’s introduction to the index, which gives a chart that points out the features of the BPI compared to RPI, KenPom, and Sagarin, and also describes the benefits of their system that they believe is the most accurate assessment of team rankings. ESPN notes that their numbers include details that are “pretty technical and many people won’t be interested, so we won’t go into detail, but we think they improve how the tool works.” Considering the great technicality with which many purists understand Sagarin and KenPom, it would actually be quite useful to release this ‘technical’ information for comparison’s sake. Regardless, the BPI appears to be quite similar to these accepted ratings. BPI accounts for pace when measuring scoring margin, it awards value to winning close games more than close losses, and it includes detailed strength of schedule numbers.
The one glaring original feature of the BPI is that it de-values games played by teams missing key players. A game’s weight is decreased by 15% anytime that team plays without one of its top five most significant players, measured by minutes played. This is a fun new feature that addresses the element of injuries and suspensions, but it’s immediately a spot for skepticism. This is a very basic way of weighting teams’ full strength ideals. Are minutes per game alone representative of the most valuable players? And why not give each individual player a ‘value’ weight based on minutes played, instead of just giving each of the top five players an equal weight and cutting it off there? The number implies that Creighton missing its star Doug McDermott is the same as it missing role player Jahenns Manigat, and it also suggests that Syracuse be de-weighted for missing Scoop Jardine but not for missing Brandon Triche, who happens to be sixth on the team in minutes played. This is all very subjective and needs obvious tweaking, meaning that a newer and improved rating will soon be next, so why fall in love with this one?
The early returns on the BPI also show that the initial rankings are incredibly similar to the KenPom ratings. One observer, John Templon (@nybuckets), measures a .989 correlation between the two ratings in terms of comparing where a team is ranked in each list. If you’re already so fond of the Pomeroy numbers, and this is so similar, then why even check it out? On the other hand, the BPI will obviously be seen by an exponentially larger audience given the Worldwide Leader’s reach, and if it is so close to the accepted advanced rankings then it will make a positive mark on fans. There’s definitely no downside to spreading the culture of predictive ratings that go above and beyond the practically irrelevant AP/ESPN Top 25 Polls.
We will take a wait-and-see approach with the new BPI. For now, the rankings will most certainly go into our bookmarks alongside the other advanced rating systems for reference. But it’s still unclear how ESPN will pimp its new system, where and when they will refer to a team’s ranking, and what the reaction will be from casual fans. We don’t love the idea that ESPN is essentially stealing the KenPom recipe, tweaking one or two ingredients, and spitting out something to call it brand new, but that of course is what any business idea is all about. They are trying to establish credibility with an idea that has already been explored and conquered by other statisticians. Credit goes to ESPN, though, for referencing KenPom and Sagarin throughout their introduction so as to not act like they’re the first to come up with such a system. We see an overall positive coming from this unveiling, since it will open many fans to this kind of analysis, but we’re just not sure if it’s all that groundbreaking or relevant in the big picture.