Monday, March 26, 2012
Pre-tournament BPI and the Final Four
By Alok Pattani
The "BPI bracket" -- which had all four No. 1 seeds making the Final Four -- took some big lumps over the weekend, with both Syracuse (No. 2 in pre-tournament BPI) and North Carolina (No. 5) losing in the Elite Eight.
In the ESPN.com Tournament Challenge, this bracket currently has 700 points and ranks in the 59.95th percentile. Projected national champion Kentucky is the only team BPI still has left in the field.
BPI was not designed to be predictive of tournament games, but rather to rank the overall quality of team’s résumés leading up to the tournament. With that said, the previous five years of BPI data was used to design a tournament prediction system that has done fairly well in the past; for more on that, see here.
Other statistical systems were higher on the No. 2 seeds that advanced -- Ohio State and Kansas -- but BPI had both teams ranked one spot behind the opponents they defeated (Ohio State was third, Kansas was sixth). And of course, that was before accounting for the ineligibility of Fab Melo and the injury to Kendall Marshall, which no major statistical system incorporated into its tournament predictions.
If you look at the average pre-tournament ranking of the four teams that made the Final Four according to BPI and some of the other systems, BPI actually has the lowest average among these five notable ranking systems.
• BPI: 5.3
• Sagarin: 6.3
• RPI: 6.8
• KenPom: 6.8
• LRMC: 7.3
(One key differentiator for BPI was having Louisville 11th before the tournament, the Cardinals' best ranking among these systems.)
Another way to look at this is to consider the BPI-based percentage chances of each team getting through to each round of the tournament. According to BPI, the Final Four is made up of teams that were all among the 10 most likely candidates to earn a spot in New Orleans.
Notably, BPI gave Kansas a solid 36.5 percent chance of making the Final Four before the Round of 64. That was the fourth-highest in the field and just behind North Carolina at 37.2 (again, with no anticipation of the Kendall Marshall injury).
The BPI-based system also gave Louisville a reasonable 10.0 percent chance of making the Final Four, the highest of any team not seeded No. 1 or No. 2. For comparison’s sake, KenPom’s system gave the Cardinals just a 4.7 percent chance of making the Final Four, and only 6.8 percent of ESPN.com Tournament Challenge brackets picked Rick Pitino’s squad to make it to New Orleans.