BCS to no longer include margin of victory

[BobbyBlueChip]

Well, I guess Nebraska's second stringers may even play a down against Troy State and Utah State. But I bet the Longhorns will still try to hang a half-a-hundred on North Texas, Houston and Tulane.


Several pollsters expected to leave BCS

--------------------------------------------------------------------------------
Scripps Howard News Service


The Bowl Championship Series -- college football's lightning rod for controversy since its inception in 1998 -- is veering off in yet another direction on its elusive search for the perfect formula to provide a No. 1-vs-No. 2 national title game.


BCS coordinator Mike Tranghese (also commissioner of the Big East Conference) has notified the founders of the eight computer rankings that make up one fourth of the BCS formula that in order to be included in this year's rankings, the computers must no longer use margin of victory in their formulas.


At least one of the computer experts -- David Rothman -- has already declined to change his formula. Herman Matthews, founder of the Matthews/Scripps Howard rankings -- has followed suit.

And according to Matthews, two others -- Jeff Sagarin of USA Today and Peter Wolfe of the Los Angeles Times -- are unlikely to change their formulas as well.


"I just don't think I'm going to change my rankings,'' Matthews said Thursday after receiving the following e-mail from Tranghese:


"My apologies for the informality of this writing, but I am in San Francisco at the CCA (Collegiate Commissioners Association) Meetings and wanted to get this to you in a timely manner.


"The BCS Board has asked me to extend to you the opportunity to continue as part of the BCS poll.


"One condition of your continuance in the BCS will be the elimination of Margin of Victory in your calculations.


"Please inform me by no later than 3:00 p.m. on Monday, June 24 of your willingness to continue with the BCS without the use of Margin of Victory as a component ... '


Rothman was less than diplomatic in his response.


"Sorry, but I don't sign my name to something which isn't the very best I can do. Not only do I use margin of victory to some extent, but the BCS should accept no computer system that would give a tie in the case of:

A 10, B 6.


B 9, C 6.


C 7, A 6.'


"Since most humans polled would agree that the correct ranking is A, B, C in this test. The current efforts to remove margin of victory are only going to degrade the computer's image, as human participants will continue to do the best they can, utilizing all information that can be ascertained, including margins.''


Three computer rankings do not take into account margin of victory -- Seattle Times, Richard Billingsley and Wes Colley -- while Ken Massey has agreed to change his formula to eliminate margin of victory. All four will apparently remain in the BCS formula.


Started in 1998 by SEC commissioner Roy Kramer, the BCS has produced four consecutive undefeated national champions (Tennessee, Florida State, Oklahoma and Miami). But the last two years the system has been mired in chaos over who should be playing the top-ranked team in the BCS 1-2 national title game.


In 2000, Florida State (11-1) got the nod to play Oklahoma in the Orange Bowl despite Miami (10-1) beating the Seminoles during the season. The Sooners' 13-2 victory did little to quell the controversy, fueled by many coaches who believed the Seminoles' larger margin of victory landed them in the title game.


Last year a string of late-season upsets led to the closest final BCS standings ever. Nebraska (11-1) got an invitation to the Rose Bowl by a tiny margin over Pac-10 champion Oregon (10-1) and Colorado (10-2), a team that routed the Cornhuskers, 62-30, in the regular season and claimed the Big 12 championship.


Miami's 37-14 romp over Nebraska in the Rose Bowl only solidified the opinion of critics who said the 'Huskers had no business playing in a national title game when they couldn't even win the Big 12.


Matthews said removing margin of victory from his rankings produced even more statistical nightmares. He ran his formula -- minus margin of victory -- after this year's bowl games. Miami remained No. 1 but Tennessee, No. 2 prior to the bowls, fell to No. 3 despite a 45-17 victory over Michigan in the Citrus Bowl. The bowl loss actually moved Michigan up in the amended rankings from 20th to 18th.


"It is clear that such a move lowers the accuracy of the rankings as well as predictions,'' Matthews wrote in a response to Tranghese. "It is surprising that you would bow to the demands of a public that knows so little about these matters and has reacted to a great extent on emotions fueled by perceived injustices.


"The concern of coaches who wish to avoid running up scores is commendable but unnecessary. Virtually all systems using MOV also use a diminishing returns principle that prevents excessive increase of ratings in such cases.''

Matthews later pulled out of the BCS rankings.

 

[JJP]

IMO, this completely makes the BCS formula worthless. Power Ratings by every criteria I've ever heard of are based on two main factors: point differentials and strength of schedule. In the example below, who is the best team:

Team A: 40 Team D: 0
Team B: 28 Team D: 20
Team C: 14 Team D: 13

Its pretty obvious to anyone that Team A is the best team, yet according to the new BCS formula, Team C is rated equal to Team A. I thought the whole point of adding the BCS was to cut down on human error (i.e. the polls).

 

[BobbyBlueChip]

Actually, the more I've thought about this, the margin of victory isn't all that significant in these polls anyway.

Most of the remaining systems use some sort of rushing average for/against , passing yards for/against, etc., and team A would most likely have the best numbers in their game against Team D and be ranked ahead of Team B & C

 

[IE]

for background on this the collegebcs site had this :

============


The BCS formula takes four basic factors:

Number of losses
Average rating in the AP and Coaches' polls.
Average rating in the eight selected computer ranking systems (after throwing out the best and worst ratings).
Strength of Schedule ranking divided by 25.
Those four numbers are added to get the basic BCS rating. Lower numbers are considered better.

New for 2001 is a "quality win" component. A team will get a bonus point reduction based on a sliding scale for each win over a top 15 team, with one illogical exception. A win over the #1 team in the basic ratings earns a 1.5 BCS point reduction, a win over #2 a 1.4 point reduction, and so on down to a 0.1 point reduction for a win over #15. The quality win bonus is then subtracted from the basic BCS rating and that gives the final BCS rating. If the QW bonus changes the order of the teams in the rankings, they do not go back and recalculate the bonus again and reapply it. Obviously, you could get into an infinite loop doing that sort of thing. The bonus is only calculated once, based on the basic BCS rating.

The illogical exception is that bonus reductions will not be awarded in the case of a conference championship game that is a regular season rematch and the team that won in the regular season wins again. In short, if one team beats a top-15 team twice, it only gets quality win credit once.

Other than that exception, all games count in all parts of the BCS formula.

The QW component, like every other part of the BCS, is based on a team's rating at the time of the calculation, not the time the game was played. It has to be that way because there are no official BCS ratings until mid-season. If they used game-time ratings, there could be no QW bonuses for the first half of the season. Obviously, that would be ridiculous.

Also, since the QW component is recalculated week-to-week, it could change as the rankings of a team's opponents change. This means if your team beats a highly rated team early on, but that team goes on a losing streak and drops out of the top 15, the bonus for beating them goes away. Then, if they pull a North Carolina and get good again, working their way back into the top 15, the the bonus would reappear.

Here is an example of how to calculate the BCS ratings:

Team A is 10-1. That is 1 BCS Point for the loss.
Team A is rated 3rd in the AP poll and 4th in the coaches poll. That is 3.5 BCS Points
Team A is rated 1st, 1st, 3rd, 5th, 2nd, 4th, 9th, and 3rd in the 8 computer rankings. Throw out the best (1) and worst (9) and average the others for 3 BCS Points.
Team A has the 52nd rated Strength of Schedule in Division I-A. Divided by 25 gives 2.08 BCS Points
Team A's BCS rating would be 1 + 3.5 + 3 + 2.08 = 9.58 Points.
Team A has among its 10 wins victories over the #6 and #8 teams in the base BCS ratings. That results in a reduction of 1.8 BCS points (1.0 for beating #6 and 0.8 for beating #8). This makes its adjusted BCS rating 7.78. If, for example, Team A has a win over #6 and two wins over #8, its adjusted rating would still be 7.78. They do not get credit for beating #8 twice, just once.

Which computer rankings are being used?
Seattle Times (Jeff Anderson/Chris Hester)
Richard Billingsley
Wes Colley
Kenneth Massey
Scripps Howard (Herman Matthews)
David Rothman
Jeff Sagarin
Peter Wolfe
What do you know about the different computer rankings?
Not a whole lot. Most of the formulas are proprietary. Some are more forthcoming about what goes in than others. All of the systems use the same basic set of data (except where noted): Date of game, location of game, score. What distinguishes them is what they do with the data, how much they weigh certain factors, and what set of teams they rank. Unless otherwise noted, all publish ratings from the beginning of the season and therefore have some prior season bias at least early on. In those systems, at some point, the prior season data is no longer relevant and each season stands on its own.
Seattle Times (Anderson/Hester)
Rates D1A teams only. Does not consider margin of victory. Factors in conference strength, which is based on how conference teams do in non-conference play. Does not publish until after 5th week.

Richard Billingsley
Rates D1A teams only. Ratings system rewritten for 2001 to remove margin of victory. Carries a team's rank over from previous year and values early part of season more highly. Has a detailed explanation on his site.

Wes Colley
Rates D1A teams only, plus provisional 1A teams (like Troy St in 2001). Publishes his formula on his website. Does not use margin of victory. Publishes ratings at the beginning of the season, but uses no prior season data. Everyone starts at 0.5.

Kenneth Massey
Ratings system rewritten for 2001 to remove margin of victory. Details not yet available.

Scripps Howard (Herman Matthews)
Rates the same set of teams Sagarin does. Uses margin of victory, but beyond that, not much is known about this system.

David Rothman
Rates all NCAA and NAIA teams. Will give you a copy of his programs if you want. I have them, but have not had a chance to read through them yet. Minimizes blowouts and does not weigh recent results more heavily. Does not publish until sometime in October.

Jeff Sagarin
Rates 241 teams (all DI, plus a few DII teams). His ratings have a starting point, which is factored out about halfway of the season. The ratings are designed to be a score predictor. Weighs all games equally regardless of when they are played.

Peter Wolfe
Rates all NCAA and NAIA teams. Has a hard cap of 21 points on margin of victory. Does not publish rankings until October. Rankings based on actual outcome vs predicted.

 

[Valuist]

I've heard Sagarin interviewed and there's no question margin of victory is a main component in his rankings since point differentials play such a large role. I've also seen Massey's ratings and they are very similar to Sagarin; his numbers will be completely worthless now that he is removing margin of victory.

I make my own power ratings and yes, without margin of victory, it would be impossible to rate teams. It comes down to the power rating of the opponent, then adding (or subtracting) the point differential from the game, then factoring in (or out) for home field advantage. I blame the Miami people for all this. Since they claim they were "screwed" back in 2000, the BCS decided to rework their formulas to see how they could've gotten Miami in.

The obvious answer? Just have a playoff and they never need to worry about this nonsense.

 

[BobbyBlueChip]

I stand corrected after reading IE's article. It doesn't make any sense at all now.

 

[IE]

BCS makes a few minor tweaks to its formula
By RICHARD ROSENBLATT
AP Football Writer
June 25, 2002


NEW YORK (AP) -- Once again, college football has changed its formula for determining which teams play in its national championship game.

The Bowl Championship Series on Tuesday eliminated the margin of victory calculation used in the computer ratings portion of the BCS standings, and also reduced the number of computers used from eight to seven.

``By eliminating margin of victory, it takes the idea of how you win and puts it in a proper perspective,'' said Big East commissioner Mike Tranghese, who serves as the BCS coordinator this season. ``The coaches didn't want it and the athletic directors didn't want it. We're putting more value in the strength of schedule.''

In addition, teams will receive bonus points for wins over opponents in the BCS' final top 10 instead of the final top 15 as they did last season.

The margin of victory change is not a drastic measure, but it shook up the computer ratings lineup. Out are Herman Matthews and David Rothman, both of whom decided not to eliminate margin of victory from their ratings. The New York Times poll is in after a year's absence.

``They told us they'd remove the margin of victory element and that they wanted consideration,'' said Tranghese, adding that at least nine other computer services also made pitches. ``They've been in it before, and have been good partners.''

The computer ratings count 25 percent in the BCS standings -- the other elements are The Associated Press media poll and the USA Today/ESPN coaches poll, strength of schedule and won-loss record.

If the margin of victory change had been in effect last year, there's a good chance that Oregon -- not Nebraska -- would have played Miami for the national title in the Rose Bowl.

In four of the eight computers that did not use margin of victory in 2001, Oregon was second in one and third in the others. In the four that did, Oregon was sixth in one, seventh in two and eighth in the other.

Oregon finished No. 2 in the media and coaches' polls, but fourth in the BCS standings.

``It's possible there could have been a different team in there last year,'' Tranghese said. ``It was a very close call and if I recollect, Oregon won about a half dozen games by less than a touchdown.''

The 2000 season ended with Miami believing it should have played Oklahoma for the title instead of Florida State after beating the Seminoles in the regular season. Before the 2001 season opened, the BCS added a bonus-point plan that would have put Miami in the 2000 title game if it were in place at the time.

Tranghese and his fellow commissioners whose conferences are in the BCS also discussed adding an oversight committee to play a role if problems arose with the final BCS standings.

``There was genuine interest in a human element,'' Tranghese said. ``But if we are going to make a change there has to be a consensus and there was not. We thought the stakes were too high to put that in the hands of a few individuals.''

Tranghese said he's well aware the system is far from perfect.

``We'll never get a system that will satisfy everybody,'' he said. ``We'll evaluate it every year to see if it's better. Hopefully, this is a more appropriate and fair system.''

The seven computer ratings for the upcoming season are Anderson & Hester, Atlanta Journal-Constitution Colley Matrix, Richard Billingsley, Kenneth Massey, Jeff Sagarin's USA Today, the New York Times and Peter Wolfe. The team's worst rating will be discarded and the average of the top six will be used.

The first BCS standings will be released Oct. 21.

With a 12-game schedule this year instead of the usual 11, teams will need a 9-3 record and top 12 ranking in the final BCS standings to be eligible for one of the four BCS games -- the Fiesta, Rose, Sugar and Orange bowls. The Fiesta Bowl plays host to the national title game on Jan. 3.

The BCS was formed in 1998 in an effort to match the top two teams in a national title game. The Atlantic Coast, Big East, Big Ten, Big 12, Pac-10 and Southeastern conferences, plus Notre Dame, are BCS members.