Home

Fixing the Final 2003 AP College Football Poll

1) Southern Cal 12-1
2) Louisiana State 13-1
3) Oklahoma 12-2
4) Ohio State 11-2
5) Miami (FL) 11-2
6) Michigan 10-3
7) Georgia 11-3
8) Iowa 10-3
9) Washington State 10-3
10) Miami (OH) 13-1
11) Florida State 10-3
12) Texas 10-3
13) Mississippi 10-3
14) Kansas State 11-4
15) Tennessee 10-3
16) Boise State 13-1
17) Maryland 10-3
18) Purdue 9-4
19) Nebraska 10-3
20) Minnesota 10-3
21) Utah 10-2
22) Clemson 9-4
23) Bowling Green 11-3
24) Florida 8-5
25) Texas Christian 11-2

To the left is the final 2003 AP college football top 25. The fixed final AP top 25 follows the article below.

This is the first AP poll I've fixed thus far with a problem at the top of the list. BCS champion Louisiana State played 5 AP top 25 teams and defeated four of them. The AP poll's #1 team, Southern Cal, played 2 AP top 25 teams and defeated them both, but took their loss to an unranked opponent.

How many national champions have played as few as 2 rated opponents? Southern Cal 2003 is the last one to do so. Since that time, every national champion has defeated at least four rated opponents. Kind of like LSU in 2003. The last national champion to defeat fewer than four was Michigan 1997, but that year was similar to 2003, as the national co-champion and AP #2 team, Nebraska, did defeat four rated opponents.

To find a national champion that played a weaker schedule than USC 2003, we have to go all the way back to 1984, when Brigham Young played no rated teams and became the worst "national champion" of all time. In fact, I do not personally recognize BYU '84 as a national champion at all.

I do, however, recognize Southern Cal as a national co-champion of 2003. The question is, can they logically, validly be ranked #1? LSU obviously played a tougher schedule. And they won their games by a slightly higher average margin of victory (22.9 to 22.7). Case Closed? Almost, but not quite...

LSU running over Oklahoma for the 2003 BCS national championship

13-1 LSU ran over Oklahoma 21-14 for the 2003 BCS national championship. But Southern Cal held them off for the AP poll's top spot, winning the Rose Bowl 28-14 over 10-3 Michigan.

Who's #1?

2003 was the year the BCS broke. At the end of the regular season, Southern Cal was #1 in both major polls, and Louisiana State was #2. But the BCS, of course, had their own rankings, based on an infamously convoluted (read: bad) formula, that had Oklahoma #1 and LSU #2 (by a whisker over USC). Supposedly designed to match up the top two teams in the country, the BCS delivered instead #2 vs. #3. In the aftermath, we ended up with a split championship, and the BCS went back to the drawing board and revised their ranking formula (it is still bad, but not nearly as bad as it was).

I believe AP voters had a strong incentive to keep USC #1 after the bowls out of a sense of fairness. Because they could not play each other, USC and LSU should be considered co-champions for 2003. But the only way that was going to happen was for the AP poll to go for USC. I am certain the coaches poll went for Nebraska over Michigan in 1997 for the same reason. And I don't think there is anything wrong with that. If I had been an AP voter in 2003, I would have voted USC #1 for that very reason, even though I think LSU is the best logical choice for #1. It's just too bad the AP poll couldn't extend the same courtesy to Auburn in 2004 (they have a much better national championship resume than USC 2003 does, including 5 wins over ranked opponents). Or Penn State in 1994.

I also think AP voters had a strong incentive to give the BCS the proverbial finger.

Of course, none of this would have been an issue had LSU and USC played each other like they should have. After Oklahoma got whipped 35-7 by Kansas State in the Big 12 title game, there would have been no controversy had they been left out of the national championship game. Computers don't bitch and moan. Writers, however, make a living at it. Which brings up the question...

Why Was Oklahoma #1 in the BCS Rankings?

The BCS formula had been a train wreck waiting to happen from the time it was devised. In 2003, it gave computer rankings equal weight with human polls, and included redundant factors that gave teams extra points based on their strength of schedule, their straight record, and wins over "quality opponents" (teams ranked highly in the BCS). By redundant, I mean that those factors should have already been accounted for in both the computer and human rankings. And each was a poor measurement in and of itself.

The strength of schedule (SOS) component, for example, was based on opponents' straight records (2/3) and opponents' opponents' records (1/3). And that is a poor measurement of strength of schedule, as I covered in more detail in the Strength of Schedule section of my How to Rate Teams guide. By that standard, #25 Texas Christian was a "stronger" opponent in 2003 than #6 Michigan.

In 2003, Oklahoma had a big lead in the BCS formula's SOS component. Why? Almost entirely because they played 9-4 North Texas and 9-5 Fresno State, both unranked. Now, it is true that those teams are far better than LSU opponents Western Illinois and Louisiana-Monroe, for example. But it is also completely irrelevant, because neither North Texas nor Fresno State would have been any threat at all to LSU (or USC). North Texas was 0-4 against teams that finished with a winning record, and Fresno State was 1-5. A 9-4 Sun Belt team is not the same as a 9-4 SEC team (as one example), yet this component treats them the same. Oklahoma beat North Texas 37-3 and Fresno State 52-28, but they only beat 4-9 Alabama 20-13! That says it all right there.

And then Oklahoma won the computer rankings, which at that time were given equal weight with the human polls. You can read my full analysis of computer rankings here. But the easiest way to see the problem with computer rankings is to simply look at who computers have chosen as #1 throughout history. Ohio State over Tennessee in '98, Florida State over Florida in '96 (and over Nebraska in '94, Alabama in '92, Miami in '87), Notre Dame over Miami in '89, vice versa in '88, Oklahoma over Penn State in '86, and on and on. A parade of the ridiculous, and lots of teams rated right on top of the teams that beat them.

After 2003, the redundant components were discarded, and computer rankings were given less weight than human polls (which is a good first step to an even better fix... dumping them entirely). Furthermore, BCS computer rankings are no longer allowed to account for margin of victory. Which is yet another bad idea, but I have tarried long enough...

Louisiana State vs. Southern Cal

So as I covered in the intro, LSU had a far tougher schedule than did Southern Cal in 2003 (by the AP poll's own rankings). And Oklahoma was a better opponent than anyone USC faced (as even the AP poll itself agrees). And LSU was #1 in defense (yards per game), and #31 in offense, while Southern Cal was #30 in defense, and #14 in offense. And LSU had a slightly higher average margin of victory. Did AP voters have any reason to vote USC #1 other than a sense of fairness regarding the national championship, and giving the finger to the BCS?

Yes, they did. Along with those political factors, USC ended up #1 in the AP poll simply because LSU took their loss on October 11th, two weeks after Southern Cal took theirs. That put USC ahead of LSU to stay in the next poll. AP voters rarely demote a team that wins. When they do, it's because that team performed poorly in a close game. But Southern Cal cruised through the rest of their (weak) schedule with ease, giving LSU no chance to pass them up. In fact, LSU had a close (17-14) game with #13 Mississippi on November 22nd. I would not call that a poor performance, but USC had no close games at all other than their triple-overtime loss to unranked Cal.

So despite LSU's higher average margin of victory, which is the product of their stomping on their patsies more than USC stomped on theirs, the AP poll has a case for USC built on performance. Again, USC had no close games other than their loss. LSU, on the other hand, had 3 close games other than their loss (the Mississippi game, 21-14 over #3 Oklahoma, and 17-10 over #7 Georgia). Is that enough?

That is a tough one. I mean, for me, it's a no-brainer. LSU #1. What's tough is determining whether or not USC is also a valid choice. USC did perform better, but not much better, and they did it against a schedule that was less than half as tough! And everything else points to LSU. I've been mulling this over for a great deal of time now. I bend over backwards to give AP voters their way as much as possible, but this one is breaking my spine. So we're going to have to compromise.

The fix here is sharing the #1 slot between USC and LSU. The AP voters and common sense both win! My spine feels better already. On to the rest of the poll, which requires a lot of fixing this time around. Which only goes to show that the AP voters don't have any idea what they're doing, and I probably shouldn't have taken their USC choice as seriously as I did...

Michigan

So Michigan (10-3, #6) beats Ohio State (11-2, #4) 35-21, thereby winning the Big 10 and securing a higher rating than Ohio State. Their reward? Facing Southern Cal, whom the AP poll itself feels is the best team in the land, predictably losing 28-14 (the closest anyone but Cal got to them), and dropping back behind Ohio State, who beat Kansas State (11-4, #14) in the Fiesta Bowl by 7. A team that the AP voters apparently think Michigan would beat too (being #6, and KSU #14). And these are the people I'm supposed to trust on who's #1?

Michigan to #4, Ohio State to #5, and Miami (FL) to #6. On a side note, I would rank Georgia #3 or 4 myself, but not doing so is a valid choice too.

Miami vs. Miami

Originally #5, this AP poll ranking is 11-2 Miami-Florida running on the gas fumes of its reputation as a juggernaut (just one year prior). What no one knew then was that Miami was on its way down, down, down. Of course, what happened in subsequent years, while interesting, is irrelevant. What matters are the facts of 2003. Which are these. Miami beat a couple of unranked teams at home by 2 and 7 points, lost 31-7 to unranked Virginia Tech, and lost 10-6 to #15 Tennessee (at home again). On the plus side, they beat #11 Florida State twice and #24 Florida once (though barely, 38-33 at home).

#7 Georgia (11-3) lost to Florida 16-13, but they beat Tennessee 41-14. And that Florida game was their only upset loss, half as many as Miami had. And they beat as many rated opponents. And they had only one close win over an unrated opponent. This is no contest.

Then there's #8 Iowa (10-3), who has 2 upset losses, just like Miami. But they beat 4 rated opponents (one of which is now-#4 Michigan), one more than Miami. They had one close win over an unrated opponent, half as many as Miami. And they stomped on Florida in the Outback Bowl 37-17 (the team that lost 38-33 at Miami). Also no contest.

Which brings us to the main event. Miami vs. Miami. Mano a mano.

The Main Event

Miami of Ohio, or "Little Miami," was 13-1 and ranked #10. They lost only to #8 Iowa in their opener, and twice defeated a rated opponent, although it was the same opponent each time: #23 Bowling Green. Their bowl game was a fairly worthless 49-28 win over unrated 9-4 Louisville. Little Miami had 2 close wins over unrated opponents, the same as Miami of Florida (or "Big Miami").

Like the AP poll voters, I have a hard time believing that Little Miami was better than Big Miami. Of course, we all had a hard time believing that Boise State was really better than Oklahoma in 2006 and that West Virginia was better than Georgia in 2005, at least until those games were actually played. So let's ignore all those future NFL players for Big Miami and look at the facts.

In Big Miami's favor, other than the fact that we all simply know they were a better team, is their tougher schedule. They played and defeated one more ranked opponent than did Little Miami, and while the Big East was weak in 2003, it was better than the MAC.

In Little Miami's favor is that inexcusable 31-7 loss Big Miami took to unranked Virginia Tech. And Big Miami had two losses to lower-rated opponents, Little Miami none. And Little Miami won their two games against rated Bowling Green 33-10 and 49-27, while Big Miami's 3 wins over rated opponents were all by a touchdown or less. And lest you think that Bowling Green was just overrated at #23, the opposite is true, and in fact Bowling Green will end up very near the top ten in the fixed AP poll (details below). All 3 of Bowling Green's losses were to ranked opponents, including a 7 point loss to now-#5 Ohio State. And they defeated two Big Ten teams: Northwestern in the Motor City Bowl and #18 Purdue. They also beat 10-2 Northern Illinois 34-18.

So the facts line up for Little Miami here. In the end, I asked myself this question: If Big Miami had played Little Miami's schedule and performed exactly the same, and vice versa, where would these teams be ranked? And that's a no-brainer. Big Miami would have finished at #3, and there would have been even more whining about the BCS that year.

So we are holding Little Miami's glove up high and declaring them the winner here. Move Georgia to #6, Iowa to #7, Little Miami to #8, and Big Miami to #9. At this point, you may have noticed that I have "forgotten" 10-3 Washington State, who had been ranked at #9, one spot ahead of Miami of Ohio...

Washington State

Washington State's whole season boils down to a big 28-20 Holiday Bowl win over #12 Texas, their only win over a rated opponent. They lost to two unrated opponents, 5-7 Notre Dame and 6-6 Washington. Florida State has the same 10-3 record, but they have one less loss to lower-ranked opponents, and one more win over ranked opponents. Also, FSU's one upset loss was to a ranked team (#22 Clemson), whereas WSU's 2 upset losses were both to unranked teams.

So we'll move FSU to #10 and drop WSU to #11, one spot ahead of their Holiday Bowl victim, Texas. You might think that would keep WSU from dropping further, but the next few fixes involve a number of teams jumping over both WSU and Texas, so drop further they shall...

Florida

Florida is ranked at #24, largely because of their 8-5 straight record. But that straight record is the product of an incredibly difficult schedule. All five of their losses are to AP top 25 opponents, and they defeated national co-champion LSU 19-7 and now-#6 Georgia 16-13. Those wins blow away WSU's one win over Texas. In addition, Florida beat 9-4 Arkansas, who sits just outside the top 25, and will be sitting inside it by the time the AP poll is repaired (details below). Given that WSU lost to two unranked opponents, and Florida none, there is no logical reason at all for WSU to be ranked higher than Florida. Yes, WSU has 2 fewer losses, but that is because they played 2 rated opponents, while Florida played 7...and you can make it 8 when Arkansas enters.

Florida did lose to Mississippi (#13, 10-3), who now sits 2 slots behind Washington State, but Florida's relevant record is better than Mississippi's anyway. Mississippi lost to 2 unrated opponents (Memphis and Texas Tech), while Florida lost to one team that is rated lower than Mississippi (#15 Tennessee). Add Mississippi's head-to-head win over Florida into the mix, and the two teams are even. Add Florida's 2 upset wins over top ten teams (LSU and Georgia), and Florida finishes an effective 2 games ahead of Mississippi. Again, the only reason Mississippi is 10-3 and Florida is 8-5 is because Ole Miss played 2 rated opponents and Florida played 7.

The fix here: move Florida all the way up to #11, dropping WSU to #12, and of course also dropping all the teams that were between WSU and Florida one spot each.

Purdue

Purdue, 9-4 and originally ranked #18, also had a better season than Washington State. Like WSU, Purdue only defeated one ranked opponent, but Purdue's win was 27-14 over now-#7 Iowa, better than WSU's win over Texas. And Purdue had only one loss to a lower-ranked opponent, one better than WSU. And Purdue's upset loss came in their season opener, and to a ranked opponent (now-#24 Bowling Green), whereas WSU, again, lost to 2 unranked opponents. So Purdue moves on up ahead of WSU too.

And in fact, since their relevant record is the same as Florida's, and the AP poll had them ranked higher than Florida to begin with, we'll move Purdue back ahead of Florida too. That puts Purdue at #11, Florida at #12, and of course all the teams that were between Purdue and Florida drop a spot too. This ranking fits Purdue pretty well, as they had overtime losses to both #5 Ohio State and #6 Georgia at the end of the season, indicating that they were nearly as good as those teams.

Bowling Green

In every year's AP poll, you may have noticed that I seem to be picking on non-BCS schools. Brigham Young, Texas Christian, Fresno State, etc. Every year, a couple of these "little big" teams fall out of the fixed top 25, and it'll happen in this poll too. But not to Bowling Green. Bowling Green's three losses came to teams the AP poll originally ranked at #8 and #10 (they lost to #10 twice). And they defeated the AP poll's #18 team. Does that sound like a #23 team to you? That's where the AP poll had them, making them a rare case of the AP poll underrating a little big team.

With zero upset losses and a win over now-#11 Purdue, 11-3 Bowling Green also has a better relevant record than 10-3 Washington State. And Florida too. And since the AP poll had BG rated ahead of Florida to begin with, like Purdue we will move them all the way up ahead of Florida.

That leaves Bowling Green one spot behind Purdue, whom they beat 27-26 on the road in Purdue's season opener (but game 2 for Bowling Green). Should Bowling Green be ranked higher than Purdue? I would rank them higher than Purdue myself, but the AP poll actually has a case for Purdue being ranked higher. The game was by 1 point, in Purdue's season opener, and Bowling Green had a game under their belt beforehand. Purdue then played a tougher schedule and performed much better than Bowling Green on the season. So we'll let the AP poll keep Purdue ranked higher.

Move Bowling Green to #12, Florida to #13, and of course all the teams that were between Bowling Green and Florida drop a spot.

Tennessee

Tennessee is a hard team to rate, primarily because they lost 27-14 in the Peach Bowl to now-#24 Clemson (9-4). But their relevant record is 2 games better than Washington State's, and even if you count their bowl game twice, they are still one game better than WSU, so we'll move them up ahead of WSU too, putting them right behind Florida. Tennessee beat Florida 24-10 in game 3, but Florida is nevertheless one game better than Tennessee in relevant record. And that's without giving extra weight to Tennessee's ugly bowl loss.

Move Tennessee to #14, drop Washington State to #15, and of course the teams that were between Tennessee and Washington State all drop a spot. Washington State's freefall is now over. And #15 is a much more appropriate ranking for a team with their resume.

Oklahoma State

Oklahoma State finished 9-4 and sitting just outside the AP top 25 at #26. Their four losses came to teams that were originally ranked #2, #12, #13, and #19, and they defeated #14. That looks an awful lot like a team that was clearly misrated. What happened here?

Oklahoma State was ranked #21 in the final regular season poll. Then they lost their bowl game 31-28 to #13 Mississippi, and fell out of the top 25. Here is something I would like an AP poll voter to explain to me someday. How can you drop a team out of the top 25 for losing to your #13 team by 3 points? What would you expect a team ranked #14-25 to do? Lose by only 2 points? But 3 points is so many that it means you suck?

Clearly the Cowboys belong in the top 25, but where? Based on their relevant record, they should be moved up ahead of Boise State, who currently sits at #19.

Boise State has one win over a ranked opponent, 34-31 over #25 Texas Christian (11-2) in the Forth Worth Bowl. Of course, TCU falls out of the top 25 as OSU comes in, leaving BSU with no win over a rated opponent. And good riddance to TCU, too. They had, of course, no win over a rated opponent, and lost to unrated Southern Miss 40-28. Moreover, they struggled in wins over seven unranked opponents (wins by a touchdown or less). So Boise State now has one win over a near-rated opponent, and they lost to unranked Oregon State. Oklahoma State also lost to a lower-rated opponent, but their upset loss came to a ranked team (now-#21 Nebraska). OSU has a win over now-#18 Kansas State, and also had fewer close wins over unrated opponents than did BSU.

Move Oklahoma State into the top 25 at #19, and Boise State and everyone behind them drops a spot. Oklahoma State now sits behind Kansas State, whom they defeated, but KSU has a better relevant record, so can stay ranked higher.

Auburn and Arkansas

8-5 Auburn and 9-4 Arkansas, like Oklahoma State, were left sitting just outside the AP top 25 when they should have been sitting inside it. Auburn defeated now-#14 Tennessee and Arkansas, and Arkansas defeated now-#16 Texas (and lost to no unrated opponents). So like Oklahoma State, we'll move them up ahead of Boise State too (Oklahoma State can remain ahead of them, since they had more poll points than either). I myself would rate Arkansas ahead of Texas, since they beat them in Austin 38-28, but the AP poll has a case for not doing so.

They do not have a case for rating Boise State higher than Arkansas, though, and for pretty much the same reasons that Oklahoma State should be rated higher than Boise State. Move Auburn and Arkansas into the top 25 at #20 and #21, and Boise State and everyone behind them drops 2 spots.

Nebraska

While we're at it, let's move Nebraska (10-3, now #24) ahead of Boise State too. Like Boise State, they have a loss to an unranked opponent, but they also have a win against a higher-ranked opponent (Oklahoma State). Nebraska's schedule was almost as weak as Boise State's, and they got stomped by Texas (now #16) and Kansas State (now #18), but Boise State did not even play a ranked opponent, so it's hard to compare performances there. Nebraska had one less close win over an unrated opponent.

Move Nebraska to #22, Boise State to #23, Maryland to #24, and Minnesota to #25. Clemson, originally ranked #22, has by now fallen out of the top 25. But not for long...

Clemson

10-3 Maryland (now #24) defeated 9-4 Clemson 21-7. Maryland thus finished 6-2 in the ACC, one game better than Clemson at 5-3. Clemson's bowl win over now-#14 Tennessee makes them even, as Tennessee is rated well-higher than both. But Maryland still has the head-to-head edge. However, Maryland also lost to unrated Northern Illinois. And that gives Clemson a one game lead over Maryland in relevant records. Also, Clemson's late-season performance was spectacular: 26-10 over now-#10 Florida State, 40-7 over Duke, 63-17 over South Carolina, and that 27-14 bowl win over Tennessee. Looks like they became a top ten team power-wise in November. Too bad they couldn't play like that all year.

The one game relevant record lead puts Clemson at #24, Maryland at #25, and Minnesota drops out of the top 25. Minnesota was 10-3, but did not even beat a team in the "also receiving votes" section. They also lost to unranked Michigan State, and had repeated poor performances, including a 31-30 bowl win over unrated Oregon.

Fixed AP Top 25

Wow, that was a long one. I'd say that was the biggest mess I've had to clean up yet. Falling out of the top 25 were Minnesota, Utah, and Texas Christian. Replacing them were Oklahoma State, Auburn, and Arkansas, which isn't much of a surprise, given that the Big 12 and SEC had the best FBS conference records against non-conference opponents. The three teams that fall out had 4 losses to unranked opponents and no wins over ranked teams. In fact, none of the three even beat an "also receiving votes" team. The three teams that replace them had 1 loss to an unrated opponent and 3 wins over rated opponents. A big improvement there.

1) Southern Cal 12-1 Louisiana State 13-1	-0.5 +0.5
3) Oklahoma 12-2	--
4) Michigan 10-3	+2
5) Ohio State 11-2	-1
6) Georgia 11-3	+1
7) Iowa 10-3	+1
8) Miami-Ohio 13-1	+2
9) Miami-Florida 11-2	-4
10) Florida State 10-3	+1
11) Purdue 9-4	+7
12) Bowling Green 11-3	+11
13) Florida 8-5	+11
14) Tennessee 10-3	+1
15) Washington State 10-3	-6
16) Texas 10-3	-4
17) Mississippi 10-3	-4
18) Kansas State 11-4	-4
19) Oklahoma State 9-4	IN
20) Auburn 8-5	IN
21) Arkansas 9-4	IN
22) Nebraska 10-3	-3
23) Boise State 13-1	-7
24) Clemson 9-4	-2
25) Maryland 10-3	-8

OUT: #20 Minnesota 10-3
#21 Utah 10-2
#25 Texas Christian 11-2

Fixed AP Polls
Home