December 10, 2006
BCS FolliesIf you're a college football fan you've heard of the Bowl Championship Series, or BCS. This is the ranking system that determines which two teams meet in a bowl game for the national championship. It's the system that arranged an Ohio State vs. Florida matchup this year.
If you're NOT a college football fan, you may have heard of the BCS because you overhead a fan complaining about what a mess the BCS is and how it screws up the matchup each year. Why do people complain about that? It's simple: the BCS is behaving exactly the way it was supposed to behave.
Back in the way old days (1992, that kind of stuff), the college football national champion was decided by a poll. Actually there were two polls, one of the coaches and one of sportswriters, and sometimes they didn't agree, so you would have two national champions (this happened in 1991 with the Washington Huskies sharing the title with the Miami Hurricanes).
The system was obviously subjective, but most years the two polls agreed and there was reasonable consensus on who the best team was. BUT, there were two flaws that gradually became more noticeable: it was hard for one team to pass another in the polls unless the higher-ranked team lost a game, and teams tended to be penalized for losing later in the year rather than earlier.
The two are really related. Imagine that two teams finish the season undefeated, and one of them began the year ranked, say #2, and the other began the year ranked #5. Almost certainly they will wind up ranked #1 and #2 at the end of the year, but it is very unlikely that the team originally ranked #5 will be able to pass the original #2. And since the preseason rankings are somewhat random (being based on how they did last year, coach's reputation, quality of recruiting class, and other factors that mostly aren't really relevant to the current season), this isn't really fair.
The second problem was related to the first, which was that if the two best teams wound up with one loss each, the team that lost first would tend to finish ranked ahead of the other. It's a similar effect, in that after the first team loses, if it plays well thereafter it will likely climb up to be just behind the second team in the rankings, and then when the second team loses the first team would jump ahead. Again, this is somewhat random based on how the schedules are made, and didn't seem fair.
Now, college football does have a system of bowl games, which are one-game postseason contests. And if the #1 and #2 ranked teams at the end of the year always played each other in a bowl game, then the winner could be declared #1 with no complaints from anybody (except perhaps the #3 team, about which more later). But the bowl games had ties to conferences, in which the champion of a certain conference would play in certain bowl games, so a #1 vs. #2 bowl game would be a happy coincidence. The #1 and #2 teams would often play in different bowl games, both win, and remain in the same unfair preseason-ranking-and-first-to-lose-based spots.
In the mid 1990s agreements started to be worked out wherein bowls would let conferences out of their commitments if it allowed a #1 vs. #2 game to happen. This eventually coalesced into the current BCS system. The problem then became, how to decide who is #1 and #2? If both polls agreed, or at least had the same top two in a different order, it wouldn't be so terrible, although there was still the possibility that the #3 team was being treated unfairly by the rankings due to the factors discussed above. Plus, it was certainly possible for the two polls to agree on #1 but have #2 and #3 swapped, and then what?
Well, at the time various people and organizations were developing system to produce computer rankings of teams. The computer rankings were based on factors like wins, margin of victory, quality of opponent, etc (in varying degrees, down to zero, for some of the factors). Although the formulas were a bit murky, they did away with the preseason ranking problem and the first-to-lose problem.
The first BCS rankings averaged together the polls, the computer rankings, strength-of-schedule (included to encourage good teams to schedule stronger opponents), and number of losses (which was thrown in, I guess, to bias in favor of undefeated teams, although the other systems tended to factor that in also, so it wound up biasing in favor of undefeated but unappreciated teams, like those from small conferences that played weak schedules; they got zapped by strength-of-schedule anyway).
You can get PDFs of all the BCS results from this page; it shows weekly results, so scroll down to the bottom to see the final tallies. The first year, 1998, the system worked well, with the polls and the computers agreeing that Tennessee and Florida State were the top two teams (thus making the whole BCS foofaraw unnecessary, since the polls would have sufficed). In 1999 there were two major-conference undefeated teams, Florida State and Virginia Tech, who were as expected 1 and 2 in the polls and the computers, so again the system was not needed (everybody also agreed that Nebraska was #3, but Wisconsin, #4 in both polls, wound up #7 in the BCS due to bad computer rankings, weak schedule, and 2 losses). I'll point out that in both years the #1 going in (Tennessee in 1998 and Florida State in 1999) wound up winning the bowl game, so the BCS overall didn't affect the national champion.
The year 2000 was the first test for the BCS formulas. Oklahoma was the only undefeated team and the obvious #1, but while Miami was #2 in both polls, #3 Florida State was ahead of it in the computer rankings, in fact it was rated #1 by the computers. This was despite the poll-influencing fact that Miami had beaten Florida State during the regular season. But the computers had spoken, and so Florida State went off to play Oklahoma in the Orange Bowl, a game which Oklahoma won without too much trouble. So the #1 team won the championship and everybody was reasonably happy. Except people in Miami, who felt they should have had a shot at Oklahoma (Miami wound up beating BCS #7 Florida in their bowl game). People started agitating that the BCS was a bit unfair and head-to-head games should be recognized and all that.
Let's pause to consider what was going on. Miami seemed to be benefiting from voter bias. Although winning a head-to-head is a legitimate way to decide between two teams, over the course of the season Florida State had apparently been a better team, per the computers. I mean, Florida State had lost to Miami: a pretty good team, right? Miami meanwhile had lost to...well wait, who had they lost to? Well, it turns out Miami lost to the Washington Huskies, who were the #4 BCS team. So while #3 Miami was complaing that they had beaten #2 Florida State and so should be ranked ahead of them, #4 Washington could make the exact same statement about #3 Miami, which arguably should have put Washington at #2--right? I think the lesson drawn here is that there is no perfect system, and trusting the computers is as good a system as any. After all Miami did have a big BCS advantage from being #2 in the polls; Florida State had to be significantly better in the computers to pull ahead of them.
Anyway, what ACTUALLY happened the next year, 2001, is that they added a "quality win" component to the BCS. Basically you got points for beating a team that was in the top 15 of the BCS ranking, earning from a 1.5 to a 0.1 bonus for beating #1 through #15. So what happened? Well, this time it was Miami who wound up undefeated and #1, beloved by both polls and computers, and we had another dandy "Who's #2" controversy. The polls both agreed on Oregon #2, Colorado #3, and Nebraska #4. But the computers were almost unanimous that Nebraska was really the #2 team, with Colorado and Oregon basically tied for third. Colorado had two losses to Nebraska's one, but they played a very tough schedule, #2 in the country. In particular, they had played Nebraska and stomped them pretty good, 62-36. But they had somewhat unaccountably lost the season opener to Fresno State, and also lost to Texas, #7 in the final BCS. Except then they turned around and beat Texas in the Big 12 championship. So they had beaten the #2 and #7 BCS teams, which games them 1.4 and 0.9 "quality win" points, respectively, for a total of 2.3. Anyway the net result was that Nebraska wound up ahead of Colorado in total BCS points (with lower being better) 7.23 to 7.28. Second bridesmaid Oregon wound up crushing Colorado in the Fiesta Bowl, which was played before the Miami-Nebraska game, leaving them hope that if Nebraska won, Oregon might split the national championship (because after all the BCS fuss, only one of the two polls actually had its voters committed to vote the winner of the BCS title game as #1). But their hope was for naught when Miami won easily.
This was another year of controversy, with the #3 BCS team again having beaten the #2 BCS team, and the consensus #1 team remaining as the only undefeated anyway, rendering all the fuss somewhat pointless. There was also an interesting point which I, anyway, noticed, which was about Colorado's quality win points. They received 1.4 for beating #2 Nebraska and 0.9 for beating #7 Texas. But what if Texas had wound up #6 instead of #7? It was not, in fact, close--Texas was behind #6 Tennessee in the polls, the computer rankings, strengh-of-schedule, AND quality win points--but it could have happened. So now Colorado gets an extra 0.1 for quality wins and is now #2, because it has 7.18 points to Nebraska's 7.23. But hang on, Nebraska is no longer #2, right? They're now #3...so Colorado only gets 1.3 quality win points for beating them. Which puts Colorado back to #3. But now Nebraska is #2 again, so shouldn't Colorado get 1.4 quality win points for beating them? And so in an endless cycle. Help!
2002 brought a new wrinkle--a note in the BCS results that quality win points were calculated based on standings before the quality win deduction was made. OK, so somebody else noticed. The actual season was controvery-free, as Miami and Ohio State wound up as the two undefeated teams, 1-2 in the polls, and 1-2 in the computers. The only interesting thing was that Ohio State wound up winning the bowl game, so for the first time the BCS had actually produced a different champion than the old poll-only system (in the old days Ohio State would have been beholden to play the Pac-10 winner in the Rose Bowl and could not have played Miami). Which would have given Ohio State reason to complain, as an undefeated team, they they were unfairly penalized for a lower preseason ranking than Miami (they were #4, while Miami, as defending champion, began the season #1). This seemed like a year where the system worked perfectly, matching two undefeated teams who would not have met without the BCS, and producing, presumably, the deserving champion (although the Buckeyes, 11.5 point underdogs, had to squeak out a 31-24 overtime victory).
In 2003, you had a new twist: three major college teams all with one loss at the end of the regular season. The polls had it USC, LSU, and Oklahoma, but the BCS saw it the opposite way. So LSU played Oklahoma (and won), USC griped and groused about how unfair the BCS was, and much ink was spilled on how ridiculous it was that the #1 team in both polls didn't even get a chance to play in the title game.
But hang on. The 2003 season seemed like an example of the BCS working PRECISELY the way it was supposed. The rankings started out with Oklahoma above USC above LSU (there were other teams mixed in, but they filtered out as the season went on). USC lost to Cal 34-31 on Sept. 27, LSU lost 19-7 to Florida on Oct. 11, and Oklahoma lost 35-7 to Kansas State on Dec. 6. It seems clear that USC wound up ranked higher purely from being the first to lose a game. And if you want to play the "who did they lose to" game, Florida wound up as the #15 BCS team and Kansas State was #10, so Oklahoma and LSU lost to better teams. Although admittedly USC lost to Cal in overtime and Kansas State blew out Oklahoma...but still the computer rankings made it quite clear that Oklahoma and LSU were the two best teams (and they also played tougher schedules). Anyway LSU won the bowl game, but the AP voters (the ones who did not have to respect the BCS title game result) persisted in voting USC #1, so we wound up with a split title, which is what the BCS was designed to avoid. But arguably this wasn't so terrible; you had 3 top teams, two of them got to play for half the title and the other one got the other half.
But of course nobody but me saw it that way, so in 2004 the system was rejiggered so that the polls had much more weight. In other words, the system in 2003 was working as designed, de-emphasizing the polls because of the problems that led to the creation of the BCS ranking in the first place, and the response was to re-emphasize the polls, once again proving that BCS hindsight is 20-20. In 2004 there was nothing about wins or strength of schedule or quality wins, just two polls and the computer rankings, each given equal weight (that is, one-third for one poll, one-third for the other poll, one-third for the average computer ranking, with the high and low computer ranking (out of six) thrown out, just in case they diverged too far from the polls). What actually happened was that there were 3 major undefeated teams: USC, Oklahoma, and Auburn (plus Utah and Boise State, who nobody took that seriously). And Auburn did rightfully complain that the only reason they were #3 was because they had started the season ranked #17, while USC and Oklahoma were #1 and #2 in the preseason (since none of them had lost a game, the "first to lose" effect didn't come into play). But Auburn was also a solid #3 in the computer rankings, so this is a year where I felt the BCS rankings worked, in that they elevated the computer rankings so that they showed that the polls probably were correct (if Auburn had been a close #2 in the polls, depending on the details of how people voted, they probably still would have been #3 in the BCS, although I haven't gone back to see what effect their strength of schedule or quality wins would have had). Anyway USC crushed Oklahoma and the poll #1 won the national championship, as usual.
Last year was uneventful from a BCS ranking point of view; the calculations were unchanged, USC and Texas were the only major undefeated teams, they were 1-2 in the polls, and they met in a bowl game. The only surprise was that Texas actually beat USC, so for the second time (after Ohio State in 2002) a team that would have finished a clear #2 in the old system (USC would have been forced to the Rose Bowl against the Big Ten, so would not have played Texas) was able to take advantage of their shot at #1 to win the national championship.
So what is my appraisal of the BCS? Well, the part about matching #1 and #2 is good, as Ohio State and Texas will certainly agree; but generally the years where the ranking system "worked", in that it did something unexpected in choosing its #1 and #2, were the years that people complained about it the most and tweaked it to prevent whatever random thing happened from happening again. In other words, most people don't know what the heck they are talking about. What people are now calling for, especially Auburn, is a playoff, where instead of #1 vs. #2 you take the top (say) 8 teams and have quarter and semi-finals before the big game. I like to think this stems from a genuine desire to find the best team, not from gazing longingly at the money and publicity generated by the 65-team tournament that basketball plays every year. Nobody seems to notice that there is a playoff system now; it's just that it only has two teams, so there is controversy over who is #3. Do they not realize that if you have 4 teams in the playoffs, then the controversy will move to the #5 spot...and if they have 8 teams it will move to #9...and so on? I mean, when you have relative nonentities like Cal beating USC during the regular season in 2003, it's hard to argue that on "any given Saturday" (or Tuesday or whatever) one team can't beat another, and therefore deserves a chance in the playoffs.
So who knows. Maybe there will be a playoff. Or maybe there will be more tweaks next year. This year you have yet another controversy, as undefeated Ohio State is the clear #1, but there are various one-loss contenders for #2, with the actual BCS ranking having Florida beating out Michigan. Michigan already lost to Ohio State this year, a game that if they had won would have left them a solid #1, so they can't really complain; but in this case the computers actually have Florida and Michigan tied for second, with the poll then acting as the tie-breaker, and sure enough Florida is ahead of Michigan mainly because Florida lost on Oct. 14 (to Auburn) and Michigan lost on Nov. 18, so it's the "first to lose" thing, but didn't Michigan beat a better team (although Auburn is no slouch)...so maybe we should bring back strength of schedule or quality wins or something...but wouldn't it be unfair to ask Ohio State to play Michigan again, so maybe there should be something in the formula to bias against that...and on and on. As they say, be careful what you wish for, it may come true.
Posted by AdamBa at December 10, 2006 09:37 PM
TrackBack URL for this entry:
very insightful post,
The bottom line is that there's no potential for a perfect system. A playoff would take too long and schools would never agree to it, even if they had a playoff system on certain years people would argue how many teams should be allowed in the tournament and if a fringe team got in and ended up winning there would be an outcry of people on losing sides saying the champion shouldve never been in the playoff in the first place...Rick S.
Posted by: Rick S. at December 11, 2006 01:43 PM
Whatever system you use, someone is going to lose out who would have won under a different system (else it doesn't matter how you do it). That someone will scream like a stuck pig, right. In my day, there were two baseball leagues, each playing its own season and the top team in each played the "World" Series. Now some third rate team that had a season record of .516 can call itself the "World" Champions. In several recent years (though not this one), a wild card team that won nothing during the season was World Champion. It makes no sense. But it does keep the season exciting for longer than it would otherwise be.
Now if college football had an eight team playdown, what would it mean that the players were students? What does it mean today for that matter. When I was growing up, no college team played more than 9 games plus maybe a bowl. If the season gets any longer I would expect to see a player's union and a strike for a real salary. Let the colleges have real pro teams.
Posted by: marble chair at December 11, 2006 07:43 PM
A playoff is the only way to crown any champion, whether it's Little League Baseball or World Cup Soccer. Only the bowl guru's and college presidents associated with the big money making payments of bowl games see it differently.
They do it in small college, so it's no too tough of a schedule for a college kid.
It's done in Pro Football AFTER a 16 game schedule, so an extended schedule does not make college football last too long.
Let the championship rest with the last team standing, not the first team to lose.
Posted by: mudbug at December 11, 2006 08:14 PM