How the Tubershell Hides Just How Good the 2004 Tigers Were

There are three certainties in life: death, taxes, and Tommy Tuberville sitting on a two score lead. (H/T to Walt for that line.) This tendency to hold back on offense after gaining a significant lead became affectionately known as the Tubershell. With a conservative mindset and a quality defense, this wasn't actually a bad idea. Sure, it allowed for some late comebacks. Yes, it made games much more nerve-racking than they needed to be. But perhaps it's worst result was one of perception.

2004 and advanced stats

As Bill Connelly has said before, the BCS's biggest problem was that is was unable to fit three teams onto one field. That problem was never more apparent than in 2004, when USC, Oklahoma, and Auburn were all undefeated before the bowl games. Surely you know the rest of the story. USC and Oklahoma were chosen to play in the championship game while Auburn played Virginia Tech in the Sugar Bowl. USC clobbered Oklahoma and Auburn only beat Tech by 3. USC was awarded the BCS National Championship and no significant voting body (AP, Coaches', etc.) disagreed. Auburn finished 13-0, but only 2nd in the polls.

Fast forward to 2016. Advanced statistics in college football has taken off. Success Rate, Points per Play, Garbage Time, and various measures of efficiency can be used to compare teams through something other than gut feelings and eye tests. Personally, I've followed the results of these numbers since 2012 or so, and I've always wondered why the numbers don't go further back than 2005, and if they did, what they could tell us.

I've always wondered, until last week. From Connelly himself...

Using some of the methods I use for my S&P+ ratings (only, limited to just points scored and allowed), I have crafted estimated S&P+ ratings for every FBS team going back to 1890. (The short version: For this version, I compare points scored and allowed to a projected output figure based on the opponent at hand, then strap it to a bell curve...)

He was able to use these estimated S&P+ ratings to determine the best offenses, best defenses, best teams of each decade, and just the best teams overall, all since 1945. Pretty neat. Auburn ranked in the top ten of best teams in the 1910s and 1980s while the 1920 and 2010 offenses were some of the best ever.

But then he went too far. Connelly used the numbers to suggest that in 2004, the BCS was justified in sending Oklahoma to the championship game over Auburn, and that USC was better than both of them anyway. How dare he.

Just another perfect example of Bill hating on teams by making his spreadsheets hate on teams. Ask Michigan State.

But did you notice what raw data he used to make his estimated S&P+ ratings? "Points scored and allowed." Basically, margin of victory. Auburn fans caught on to this quickly in the comments, bemoaning the curse of the Tubershell and the lower MOVs it probably caused, but Connelly said there was nothing he could do about it. And he's right. We just don't have enough play-by-play data to do full S&P+ numbers for 2004.

However, I have an alternative to MOV. Over the summer of 2015, I stole an idea from Chase Stuart and his blog, Football Perspective. The Knockout. You can read more about it here and here, but tl;dr, a knockout occurs after the losing team's last offensive play while within one score (8 points). Put another way, it answer's the question, "when did the losing team last have a chance to tie or take the lead?"

I cranked out KO times for every FBS vs FBS game since 2005 and you can see the best and worst games and teams here. But I went one step further. I manually included the 2004 Auburn team. And of all the teams from 2005 to 2014, only five averaged faster KOs than 2004 Auburn. Pretty good, huh?

So, let's throw USC and Oklahoma into the mix.

The Results

First, USC.

School			Opponent	MOV	KO
USC	24	13	Virginia Tech	11	2:01
USC	49	0	Colorado State	49	44:48
USC	42	10	Brigham Young	32	32:21
USC	31	28	Stanford	3	3:30
USC	23	17	California	6	1:16
USC	45	7	Arizona State	38	50:06
USC	38	0	Washington	38	39:47
USC	42	12	Washington State	30	60:00
USC	28	20	Oregon State	8	11:01
USC	49	9	Arizona	40	35:12
USC	41	10	Notre Dame	31	27:49	MOV avg	KO avg
USC	29	24	UCLA	5	0:39	24.3	25:42	Pre-Bowl Average
USC	55	19	Oklahoma	36	47:46	25.2	27:24	Post-Bowl Average

My favorite thing right here is that Washington State never actually had possession within striking distance. The Cougars kicked off, USC scored a TD on its first drive, USC recovered an onside kick, and promptly scored another touchdown. That hasn't happened since.

Next, Oklahoma.

School			Opponent	MOV	KO
Oklahoma	40	24	Bowling Green	16	39:34
Oklahoma	63	13	Houston	50	49:45
Oklahoma	31	7	Oregon	24	39:27
Oklahoma	28	13	Texas Tech	15	12:57
Oklahoma	12	0	Texas	12	12:45
Oklahoma	31	21	Kansas State	10	12:18
Oklahoma	41	10	Kansas	31	28:31
Oklahoma	38	35	Oklahoma State	3	0:11
Oklahoma	42	35	Texas A&M	7	0:00
Oklahoma	30	3	Nebraska	27	45:26
Oklahoma	35	0	Baylor	35	32:46	MOV avg	KO avg
Oklahoma	42	3	Colorado	39	52:57	22.4	27:13	Pre-Bowl Average
Oklahoma	19	55	USC	-36	-46:14	17.9	21:34	Post-Bowl Average

On the other side of the scale, Oklahoma managed to win a game without ever actually knocking out the opponent. Texas A&M was down only seven but a Hail Mary didn't work as time expired. Man, it was close, though.

And finally, Auburn.

School			Opponent	MOV	KO
Auburn	31	0	Louisiana-Monroe	31	51:46
Auburn	43	14	Mississippi State	29	52:55
Auburn	10	9	LSU	1	0:09
Auburn	34	10	Tennessee	24	46:14
Auburn	52	7	Louisiana Tech	45	54:05
Auburn	38	20	Arkansas	18	55:46
Auburn	42	10	Kentucky	32	54:41
Auburn	35	14	Mississippi	21	13:01
Auburn	24	6	Georgia	18	39:56
Auburn	21	13	Alabama	8	16:51	MOV avg	KO avg
Auburn	38	28	Tennessee	10	16:29	21.5	36:32	Pre-Bowl Average
Auburn	16	13	Virginia Tech	3	38:07	20.0	36:40	Post-Bowl Average

That's right. Auburn's KO time, both before and after bowl season, blows USC's and Oklahoma's out of the water. The Tigers had their games "in control" about nine minutes sooner than the Trojans or Sooners on average. Though Auburn had games under control, the margin of victory doesn't reflect the same dominance, which I take as strong evidence of the Tubershell.

This isn't a knock on Bill, but in his article on 2004 he said,

For every awesome performance Auburn produced, the Tigers also had an only good performance against mediocre teams -- they beat 5-6 Arkansas 38-20 (that Arky team was actually pretty good), 6-6 Alabama 21-13, etc. And while we're at it, they struggled to put away a good-not-great Virginia Tech team (No. 14 in Est. S&P+) in the Sugar Bowl.

Sure, that 18 point margin over Arkansas could have been more, but Auburn led that game 30-0 with a minute left in the first half. In fact, that game was Auburn's fastest KO of the year!

Furthermore, Virginia Tech was the only team to play two of the top three, playing USC in Week 1 and Auburn in the bowl game. USC won by 11 while Auburn "struggled to put away" the Hokies with a three-point victory. However, Auburn actually KO'd the Hokies with a 16-0 lead in the second quarter, while USC didn't pull away until there were only two minutes left in the game.

Now, I'll be the first to admit that these numbers are incredibly simplistic. There are no bell curves, no home field advantages, no opponent adjustments. It truly is just an alternative to margin of victory. I think I could do opponent adjustments, but I'd need the KO times of every game played by the 37 opponents listed above. No thanks. You think maybe garbage time would be a better measurement for when a game was in hand? Talk to Chase Stuart.

So, with all that said, do these numbers once and for all give Auburn the upper hand in arguments over 2004? No. Have I just rendered all of Connelly's estimated S&P+ articles useless? Haha, no. Should Jay Jacobs raise another banner between 1957 and 2010? No comment.

I just know that the Tubershell is real. Very real.