Media have been using incorrect RPI data all season
Posted: Saturday March 12, 2005 8:43PM; Updated: Saturday March 12, 2005 9:58PM
Above, an example of the selection committee's team sheet for Illinois (with the game sites for Ohio State and Purdue intentionally changed as to not reveal the exact sheet). Follow the link in this story to see an enlarged version.
The RPI was wrong. Not the NCAA's version -- the secret, official RPI -- but the RPI that every media outlet quoted in stories and on television for the bulk of the season, was wrong.
An NCAA selection committee source confirmed to SI.com that the data from the "team sheet" provided to this site -- a close replication of Illinois' actual NCAA sheet -- for a story on Wednesday correctly represented the RPI figures used by the committee. It contained RPI figures for Illinois' opponents that, at the time, did not match the data being provided by either ESPN's InsideRPI or collegerpi.com, which have been the primary sources for media RPI coverage (this site used collegerpi.com, while InsideRPI provided the numbers for ESPN telecasts and studio discussions during the season).
So how did it happen?
When the NCAA announced it was altering its formula to include a home-road adjustment (weighting road wins as 1.4, and home wins as .6) before the 2004-05 season, it didn't reveal exactly how the adjustment would be calculated. There were two options for RPI calculators to follow: either 1) apply the home/road adjustment to the entire formula, which included a team's winning percentage as well as its strength-of-schedule factors, or 2) apply it only to the 25-percent portion of the formula that dealt with winning percentage.
ESPN's InsideRPI and collegerpi.com opted to compile the RPI by applying the home-road adjustment to the whole formula -- while the NCAA was actually only applying it to the 25 percent that dealt with winning percentage.
Have questions or feedback? E-mail Luke Winn.
This revelation does not in any way affect the deliberations inside the selection committee's "bunker" at The Westin Indianapolis -- the NCAA has been looking at the same RPI numbers (theirs) all season. The media and the general public, however, have been engaging in bubble debates with incorrect information. Collegerpi.com, as well as blogger Ken Pomeroy's kenpom.com ratings, have since corrected their data to follow a formula that accurately mirrors the NCAA's. ESPN's InsideRPI, meanwhile, is still sticking to the old -- and confirmed to be incorrect -- data.
The "boost" that was thought to be given to mid-major teams by the new RPI is no longer as substantial. Pomeroy, who deserves much of the credit for discovering the RPI discrepancies, noted some important shifts in the data in an interview on Saturday. In the miscalculated formula, he says, the MAC would have five teams -- Buffalo, Miami of Ohio, Ohio, Kent State and Akron -- in the top 55 of the RPI. In the corrected formula, Ohio, Kent State and Akron are all 60 or below. And Buffalo and Miami of Ohio, listed as Nos. 27 and 29, respectively on InsideRPI, are actually 43 and 44. Meanwhile, a major-conference bubble team like DePaul, which is 73 in the miscalculated formula, is actually 53 in the correct formula.
All of this has no impact in the selection room, which is using the correct RPI -- and only as an organizational tool. But it did have a major impact on the media, which, this site included, was unknowingly referencing RPI data inconsistent with the committee's.
So when you read stories or watch TV in the next two days and the RPI is mentioned, make sure you're looking at the correct information. ESPN -- the traditional home of the bubble debate on Selection Sunday -- undoubtedly will be conflicted, since its online data now does not match the selection committee's. For that reason, expect the RPI to play a much less significant role in last-minute predictions for the bracket.