Word on the street is that there’s a problem with college football’s Bowl Championship Series format. The sports media hates it, and have always hated it. Bloggers hate it, and likely will always hate it. Saturday Night Live hates it, or at least Will Forte had written one of his Weekend Update songs and needed a three-syllable subject. Two separate branches of government are really jonesin’ to fill out a bracket for the Government office pool. And apparently the coaches are considering the option of removing the only tangible voice they have in the matter. As the BCS system remains under siege during another long college football off-season, I’ll look at some of the important issues involved in determining the best system for college football.
Part II. What Exactly Are We Talking About Here? History (1998-2004)
Postseason Format #4
1998-1999 – Bowl Championship Series v 1.0
Starting in 1998, the Big Ten and Pac-10 joined forces with the other conferences to create the Bowl Championship Series. The Rose Bowl was added to the rotation for hosting the championship game. While those additions were almost guaranteed to eliminate controversy, another new wrinkle would eventually cause much, much more. The large role of pollsters – especially the working press – in determining who plays for a national championship was a point of dispute dating back to the Bowl Coalition. With the newly-formed BCS, the polls were now averaged and combined with three other criteria in determining rank: strength of schedule, losses and computer average. Strength of schedule took the record of a teams opponents and those opponents’ opponents, assigned a rank for each team and gave out points based on those ranks. Each loss added points to a team’s overall score. The computer average came from three objective rankings systems: Jeff Sagarin/USA Today, the New York Times and Anderson-Hester/Seattle Times. The computer scores were adjusted for deviation before they were averaged.
In the 1998 season, UCLA, Kansas State and Tennessee entered the final week of the season undefeated. Only the Volunteers survived unscathed, as UCLA went down to Miami and Kansas State fell to Texas A&M. 11-1 Florida State jumped to #2. #3 Kansas State was not invited to a BCS game, instead slipping to the Alamo Bowl. Tulane capped a 12-0 season at #10 and was also left out of the BCS. Clearly…more computer rankings needed to be added.
Postseason Format #5
1999-2001 – Bowl Championship Series v 1.1
The formula remained the same, but five additional computer rankings systems – Richard Billingsley, Richard Dunkel, Kenneth Massey, Herman Matthews/Scripps Howard and David Rothman – were added. Rather than the previous deviation adjustment on computer rankings, the new formula dopped the lowest rank of the eight and averaged the remaining seven. Teams now needed nine wins (as opposed to the previously-required eight) to receive and at-large BCS bid. Additionally, a provision was added to ensure that any team who finishes #3 or #4 in the BCS rankings is guaranteed a bid to one of the BCS games.* This is more commonly known as the Kansas State rule.
The irony here is that in the following season, Kansas State finished 11-1 (behind one-loss Nebraska in the Big 12 North), was ranked #6 behind two-loss SEC teams Alabama (the conference champ) and Tennessee (a divisional runner-up) and was not extended an at-large bid. Florida State, who began the season at #1, finished undefeated and played 11-0 Virginia Tech in the national championship game to minimal controversy.
Little (if anything) was changed before the 2000 season. That year, Oklahoma finished as the only undefeated team. Their opponent in the national championship game was #2 Florida State, whose only regular season loss came to #3 Miami. The Hurricanes, also finishing with one loss, finished above the Seminoles in both the AP and Coaches’ Polls, but were ranked no higher than 2nd in all computer rankings. The Seminoles were ranked #1 in five of the eight computer rankings. They had defeated their opponents by an average of 32 points, giving up no more than 27 in a single game (the high being their loss to Miami). Meanwhile, the Hurricanes gave eventual #5 Virginia Tech its only loss, won by an average of 25 points and surrendered 30 or more twice – once in a Week 2 loss at eventual #4 Washington, and once to then-2-7 Louisiana Tech at home. The Huskies, meanwhile dealt both #3 Miami and #6 Oregon State their only losses of the season, but also struggled down the stretch with middling Pac-10 teams, were dropped as low as 10 and 11 in some computer rankings and finished a distant fourth. Margin of victory and strength of schedule favored Florida State**, but when the Hurricanes, Hokies and Huskies all won their respective bowl games, backlash was due. Clearly, something had to change one year after it had actually worked.
Postseason Format #6
2001-2002 – Bowl Championship Series v 1.2
In response to previous years’ trouble, the BCS added a “quality win” criteria was added to the existing formula. The quality win factor essentially added bonus points to a team’s total for beating opponents ranked in the top 15, with the points being relative to that opponent’s position at the time of the poll (as opposed to the time of the actual game). The formula also began to phase out margin of victory as a determining factor in computer rankings. As a result, the New York Times and Richard Dunkel systems (the latter of which had been added just two years prior) were replaced in the formula by systems from Peter Wolfe and Wesley Colley. Additionally, the teams highest AND lowest computer scores (as opposed to just the lowest) were now dropped before the average was calculated.
During the season, Miami went undefeated and rose to #1. Nebraska was undefeated prior to their season-ending showdown with Colorado, but a 63-26 loss to the Buffaloes took them out of the Big 12 Championship game (and presumably, the title hunt). Colorado had already suffered a loss against Fresno State to open the season and a 41-7 thumping to Texas a month before upsetting the Huskers. Late season losses by SEC contenders Florida (to Tennessee) and Tennessee (to LSU) kept both Big 12 North teams high in the rankings, and also opened the door for 10-1 Oregon. The Ducks hadn’t faced a team ranked higher than 14 all year, but ascended to #2 in both human polls by the end of the season. Colorado and Oregon had similar resumes – scoring about 33-34 points per game, giving up about 22, each holding at least one “quality” win over a team in the final BCS Top 15 (Colorado over #2 Nebraska and #7 Texas, Oregon over #12 Washington State) and each having suffered a loss to a top ten team (Colorado to #7 Texas, a loss they avenged in the Big 12 title game, and Oregon to #9 Stanford). The difference was that Colorado suffered two losses to Oregon’s one. However, Colorado had the 2nd-toughest schedule by NCAA calculations, and Oregon had the 31st. Schedule strength, the additional quality win and the strength of the loss (reflected in computer rankings) kept the Buffaloes above Oregon despite the human polls. Meanwhile, Nebraska held a quality win over #11 Oklahoma, had a margin of victory nearly twice that of the other two teams and suffered their only loss to…#3 Colorado. The computers couldn’t take into account the TIME of the loss, and thus ranked the Huskers #2 or #3 across the board. The extra loss held Colorado back by .05 final points, and Nebraska was selected to face Miami. They eventually were crushed 37-14 by the Hurricanes while Oregon dismantled Colorado by a similar score (38-16) in the Fiesta Bowl. Outcry in favor of the Ducks started strong and continued strong through bowl season. The system had actually worked exactly as it had been set up to work, but the lack of satisfying results pointed at some flaws in the fluctuating BCS formula. Clearly, something, um, needed to be tweaked maybe?
Postseason Format #7
2002-2004 – Bowl Championship Series v 1.3
The War On Margin Of Victory continued as computer rankings that dared to do what they were told in previous years were removed from the process. Four of the existing rankings – Jeff Sagarin, Peter Wolfe, David Rothman and Herman Matthews/Scripps Howard – were told to adjust their computer formula to discard margin of victory. Rothman and Matthews declined and were removed from the process. Sagarin (now using the somewhat-infamous ELO-CHESS) and Wolfe recalculated. The New York Times rankings returned after a one-year hiatus, having also recalculated the formula to disregard margin of victory. With seven rankings systems, the computer average returned to the format of 2000, where ONLY the lowest rank was dropped instead of the highest and lowest. Additionally, the quality win component was adjusted. Now teams only received “quality win” modifiers for beating a team ranked in the top ten (instead of the top 15).
In 2002, Miami and Ohio State finished the season undefeated and were the clear choices as #1 and #2. The game itself was a feather in the cap of the BCS and its New And Improved formula.
That good fortune did not extend into 2003, one of the seasons that spurred the modern “screw the BCS” movement. With no teams finishing the season undefeated, three one-loss teams – Oklahoma, USC and LSU – were battling for the spots in the national championship game. Three other teams – Miami (OH), Boise State and TCU – finished with one loss but were ranked low in computer polls (especially the New York Times, which put them at 22, 34 and 39, respectively) and had very low ranks for schedule strength, and were thus out of the running for championship hope. Miami finished at #11 but was not extended a BCS berth. Meanwhile, both human polls ranked USC first, LSU second and Oklahoma third. Whereas the first two had lost in late September and early October, respectively, Oklahoma’s one loss came in the Big 12 Championship game in the last weekend of the year. The computer rankings, once again not registering the date of the loss as a tangible criteria, ranked Oklahoma first in all but two systems, LSU second in all but one system, and USC third in all but two systems. Oklahoma’s strength of schedule was at least 18 points better than either of its peers, and it held the only quality win of the group, having defeated #6 Texas. Instead of being a far and away #1, the Sooners finished by the skin of their teeth, just .88 points ahead of LSU and 1.04 points ahead of USC. With lower computer rankings and a weaker schedule, the Trojans were left out of the BCS championship game despite leading both polls. LSU, a consensus #2 in both the computers and the polls, went on to defeat Oklahoma claim the BCS National Championship and be named #1 in the final Coaches Poll (as per the poll’s contractual obligation with the BCS). However, when USC beat Michigan in the Rose Bowl, the AP voters, who had no obligation to the BCS, left the Trojans at #1, creating a split championship – the first in six years, and at this point, a clear indicator that more change was coming. Clearly…wait, I covered that part already.
In the six year history of the first BCS formula, the Series had provided a clear, undisputed consensus champion three times – 1998-99, 1999-2000 and 2002-03 – already an improvement over its predecessors. In the remaining three years – 2000-01, 2001-02 and 2003-04 – the non-human portions of the poll (including computer average and schedule strength) largely disagreed with and trumped the voters, producing results that didn’t agree with the pollsters’ selections. The computers themselves had performed as asked, but the resulting backlash demanded change.
Next Time: History ( 2004-2006 )
* – According to the official BCS site, this rule wasn’t implemented until a year later.
** – The official BCS site says “…Florida State earned the spot by finishing second in the BCS Standings, on the strength of victories over three ranked opponents (including two in the top ten.)” The primary statement is true, but the parenthetical is not – FSU defeated #7 Florida, #13 Georgia Tech and #15 Clemson if we’re going by the final rankings. It’s Miami (#2 FSU, #5 Virginia Tech) and Washington (#3 Miami, #6 Oregon State) that have wins over two top-ten teams. It’s possible this is a typo on the part of the BCS.