Get Connected
  • facebook
  • twitter
  • Sign In
  • Classifieds
  • Sections
Print

C-USA defends method of deciding home team

HUNTINGTON, W.Va. -- A Conference USA official said Monday the exact Bowl Championship Series formula was used to determine the host team in Saturday's championship game, and was not modified in any way.

The tiebreaker was needed when division winners Marshall and Rice finished with identical 7-1 league records and had not played each other. Rice was declared the host of the game, scheduled at noon Saturday on ESPN2.

Alfred White, senior associate commissioner in charge of football operations, said the C-USA athletic directors and presidents approved the application of the BCS formula in 2005, including the way the computer index rankings are handled.

Meanwhile, Jeff Sagarin, who programs and compiles one of the six computer indices used in the BCS standings, criticized C-USA's use of his ratings, along with a failure to contact him for advice on how to use them.

The computer rankings count for one-third of the BCS formula, with the USA Today Coaches Poll making up one-third and the Harris Poll making up the other third.

Marshall received votes in the two polls, an unexpected byproduct of its 59-28 dismantling of East Carolina last weekend. It didn't fare as well in the computer rankings, roughly 10 spots below Rice on average.

Neither team had an average in the top 25, a fate that scored Louisville and Fresno State a zero in that particular category. Both of those teams made the BCS's latest top-25 rankings, however, on strength of the "human polls."

By that barometer, it seemed to some that Marshall and Rice should score a similar zero in the computer factor. With Rice not receiving any poll votes, that would give the Owls a 0.0 composite score, compared to a 0.0041 for Marshall.

But C-USA "extended" the computer formula to base the rankings on a 1-through-125 basis, a method that gave Rice a hefty 0.628 on the computer factor - a number higher than BCS 10th-ranked Michigan State. MU's number was a lofty 0.55, but Rice's 0.78 was high enough to earn home field this weekend.

The methodology, clearly a departure from what the BCS uses to determine the top 25 teams, was not communicated in advance to the general public.

"The explanation was, at the end of the day, the computer rankings, which were heavily favored toward Rice, overcame the Harris Poll and the USA Today . . ." said Marshall AD Mike Hamrick, who was serving at Nevada-Las Vegas in 2005. "That's simply what I was told, and I was told that came from the actual BCS people themselves."

White confirmed as much Monday, and denied that the system was unfair or even changed.

"The formula was not modified," he said. "The formula was used."

Marshall led Rice in just one index, C-USA said, the Billingsley, 41-42. Those were the high numbers for both and therefore dropped. In the Sagarin ratings, Rice was listed at 54 vs. Marshall's 56, a result Sagarin himself disputes.

"I was NOT contacted at all by anyone from Conference USA nor anyone else about this," Sagarin said Monday via e-mail.

Sagarin compiles, with explanations, four different sets of numbers. The BCS uses the "PURE_ELO" number, in which only winning and losing matters. Sagarin's main ranking is simply known as "RATING," and he also uses a "DIMIN_CURVE" and a "PREDICTOR" index.

"You'll note that they evidently used my ELO ratings even though the explanation over the ratings says that the more accurate rating for predicting who will win is NOT the ELO," he said. "So Rice ends up as the home team even though I have Marshall as the better team in all three of my score-based ratings:  RATING [53 to 72], DIMIN_CURVE [61 to 74], PREDICTOR [47 to 72]."

That's not his biggest objection, though.

"The ultimate insult is they got Jerry Palm [of CBS Sportsline] - who has no connection with me or my rating system - to advise them on how to use the computer ratings," Sagarin said.

"I guess they thought it would be dumb to actually ask the programmers who created the ratings."

Contact Doug Smock at 304-348-5130, dougsmock@wvgazette.com or follow him at twitter.com/dougsmock.

 


Print

User Comments