Tuesday, May 12, 2009

National Champs Per Play

So I was having a debate with a friend about the 1994 Michigan Wolverines and whether the Colorado and Penn State losses kept that team from winning the national title. My friend said that '94 Michigan had plenty of athletes on defense and they were very close to being an elite team. I said that '94 Michigan's defense was just plain bad by Michigan standards, from the opening play of the season - a 70-yard touchdown pass by Boston College - to the end. Even if Chuck Winters would have knocked down Kordell Stewart's heave in the end zone and even if Amani Toomer would have hauled in Todd Collins perfect heave on the final drive against Penn State, this defense was not good enough to win a national title.
So, to prove my point, I spend a few minutes playing around with the NCAA's statistical database to see how Michigan's '94 defense, which allowed over five yards per play, would have compared to the teams that have won the title this decade. That little assignment turned into a survey of each of the ten teams that have won a national title in the aughts. I wanted to set the parameters for what a team needs to show in terms of yards per play gained and allowed to be consistent with a decade's worth of national champions. Behold:
Yards Gained Per Play:

2008 Florida - 7.13
2007 LSU - 5.84
2006 Florida - 6.34
2005 Texas - 7.07
2004 USC - 6.33
2003 USC - 6.49
2003 LSU - 5.89
2002 Ohio State - 5.61
2001 Miami - 6.57
2000 Oklahoma - 5.99

Yards Allowed Per Play:

2008 Florida - 4.46
2007 LSU - 4.42
2006 Florida - 4.32
2005 Texas - 4.39
2004 USC - 4.27
2003 USC - 4.41
2003 LSU - 4.02
2002 Ohio State - 4.66
2001 Miami - 3.93
2000 Oklahoma - 4.14

Yards Gained/Yards Allowed Margin

2008 Florida - 2.67
2007 LSU - 1.42
2006 Florida - 2.02
2005 Texas - 2.68
2004 USC - 2.06
2003 USC - 2.08
2003 LSU - 1.87
2002 Ohio State - 0.95
2001 Miami - 2.64
2000 Oklahoma - 1.85

Sagarin Strength of Schedule:

2008 Florida - 4
2007 LSU - 11
2006 Florida - 8
2005 Texas - 13
2004 USC - 7
2003 USC - 19
2003 LSU - 28
2002 Ohio State - 30
2001 Miami - 27
2000 Oklahoma - 14

Thoughts on the data:
  • Generally speaking, the BCS has done a reasonably good job of spitting out national champions like McDonald's spits out Big Macs. Over a wide space of time and geography, the end product has been fairly consistent: national champions who: (1) played schedules that can be described as "strong to quite strong;" (2) gained something in the neighborhood of six-to-seven yards per play; and (3) allowed something in the neighborhood of four yards per play. The BCS may not be perfect, but it has prevented a recurrence of 1984 BYU winning a national title without playing a quality opponent. (Sagarin's database at USA Today's site does not go back to 1984, but Soren Sorenson's site does and his ranking system put BYU's strength of schedule at 85.) Advocates of going back to the old bowl system might want to consider the fact that we haven't had a massively unqualified national champion since the inception of the BCS.
  • Using yards per play margins, national champions this decade can be put into three groups: three dominant champions ('08 Florida, '05 Texas, and '01 Miami), one team that had absolutely no business winning a national title ('02 Ohio State), and then everyone else. The numbers for the '02 Bucks are pretty amazing. Ohio State had the lowest yards per play and the highest yards allowed per play, all while playing the weakest schedule of any national champion this decade. There is absolutely no way to view that team as something other than insanely lucky. In other words, 2002 Ohio State should never be held out as a credible model for a team seeking to win a national title. Note to self: set Google Alert for "2002 Ohio State" in September for fisking possibilities.
  • One other issue: 2002 Ohio State is the only non-Sun Belt team to win the national title. If the Bucks were such an anomaly, then what does that say for the teams of the Northeast and Midwest as credible contenders for the national title? (One counter: Ohio State and Penn State have produced teams that looked more like national champions than the '02 Bucks. For instance, 2005 Penn State outgained its opponents by 1.89 yards per play. 2006 Ohio State outgained their opponents by 1.54 yards per play, even with the decimation by Florida in Glendale.)
  • Although few will make the case that 2008 Florida is one of the great teams of recent history because they lost a home game to Ole Miss, the yardage numbers put the Gators in elite company. Florida had the best offense of any national champion this decade, a defense that would stack up with most, and they played a very difficult schedule
  • For fun, here are Georgia's yard-per-play differentials for their three best years under Richt: 2007 - .75; 2005 - 1.45; and 2002 - .94. 2005 is the year that the Dawgs had a national title-caliber team (barely) and can consider themselves unlucky.
  • The main point of this exercise is to give us a yardstick for November when the various national title contenders have been identified. Yards per play gives us a good way to say that a team does or doesn't look like what we've come to know a national champion should resemble.

24 comments:

Klinsi said...

OK - my pet peeve has reared it's ugly head once again: even if you want to think in "decades" using the traditional paradigm of journalistic simpletons, why stop at 2000 when the BCS era is just over 10 years old? Why not use the full set (1998-2008) of BCS season data? Why not compare the BCS success in identifying a legitimate national champion, which seems well-founded (despite my dislike of it), to the (at least perceived) inadequacies of the earlier Bowl Alliance/Bowl Coalition attempts? Because they started in a year not ending in zero? Why the slavish adherence to an arbitrary division of celestial patterns in space-time?!?

Spacemonkey57 said...

You're ignoring scoring defense, isn't that a better metric than Yards per play?

Josh said...

Ugh, football isn't about statistics it's about wins. You can't possibly put the 2008 Gators in the same category as 05 Texas or 01 Miami when they LOST a game.

Also Ohio State under Tressel has never been guilty of running up the score against the inferior opponents on their schedule like 08 Florida did when they played the fucking Citadel for their last home game or like 05 Texas did on multiple occasions hanging more then 60. Ohio State also beat one of the most dominant teams this decade to win their National Title.

If you want to talk about the least deserving team it's probably 07 LSU, 2 losses to mediocre opponents before a championship game in their own backyard

Michael said...

Klinsi, I started in 2000 because that's the first year for which the NCAA's database has full stats. If I could go back to 1980, I would.

Spacemonkey, I like yards per play better than scoring stats because yards per play is pace neutral. For instance, it doesn't punish a defense that is paired with an offense that scores quickly. For instance, Oklahoma's defense was criticized last year for giving up a lot of points and yards, but they were paired with an offense that scored a ton of points very quickly. On a per play basis, they were better on defense than Texas. Thus, it wasn't a surprise when their defense performed well against Florida in the title game.

Josh, wins are a statistic, just like points and yards. They are also a stat containing a tremendous amount of noise. There has been a ton of research by baseball statheads and there is overwhelming evidence that run differentials are a better measure of a team's future performance than record. Digging deeper, underlying stats like offensive and defensive performance can be a better measure than runs. Football is no different.

Right, playing the Citadel is SO different than playing Youngstown State. Are you really going to take the position that Florida has far superior offensive numbers than Ohio State because of the plays they call for their scrubs in blowouts? If you really think that, then we have nothing to discuss because you have your head firmly planted in the sand. Or maybe you missed the the last Ohio State-Florida game.

'07 LSU was a weak national champion, but they were still significantly better on a per play basis than '02 Ohio State (and against a more difficult schedule, to boot).

Spacemonkey57 said...

What about points per play then? The OSU defense should get some credit for being on the field so much thanks to that crappy offense. Also, the SOS isn't as bad as you'd think, Ohio State did beat Pac Ten co-champion WSU and Texas Tech that season. They were hurt by missing Iowa in conference.

Unknown said...

It'd be interesting to see how these numbers compare to teams that didn't win MNCs.

For example, 2008 USC.

YPP gained: 6.6
YPP allowed: 3.6
YPP differential: 3.0
Schedule: 16

Source:

http://cfbstats.com/2008/team/657/index.html

and

http://www.kiva.net/~jsagarin/sports/cfsend.htm

Anonymous said...

Ohio State was the only team on that list to beat an undefeated defending National Champion on that list (who at the time had won 34-games in a row), and one of only a few teams on that list who went undefeated. Granted, their regular season schedule might not have been the hardest, but if you watched the 2002 NCG you know that they belonged on the field and they earned that victory by holding Miami to no gain on 4 downs from the 1 yard line to end the game.

Your stats are nice and I'm sure you think you are brilliant for analyzing them, but statistics or the ability to analyze them isn't everything.

I'm sorry but either one of the LSU NC's should be listed as the worst of the best. They didn't have to play USC in 03 and had 2 losses in 07 - both times they played a virtual home game.

Texas 05, Florida 08, and Miami 01 are all candidates for the best of the best in my book.

The bottom line is that the BCS NC Trophy is in the case in Columbus, deal with it.

Anonymous said...

Oh yeah, please be sure to let us all know when the BCS decides to start awarding the NC on a yards per play basis.

Anonymous said...

"Ohio State was the only team on that list to beat an undefeated defending National Champion on that list (who at the time had won 34-games in a row)"

What? Who do you think Texas beat in the '06 Rose Bowl? Why it was the undefeated defending National Champion that had won 34 games in a row-- Southern Cal.

chg said...

Do Ohio State fans intentionally try to live up to their stereotype? No one is saying the BCS trophy should be taken back from Columbus. For one thing, that would force some poor sap to visit Columbus, Ohio.

Given the epic fail served up by the big ten in every other high profile game this century, I understand being a little sensitive to criticism of the championship team. However, the numbers simply confirm what most cfb fans outside of the upper Midwest already believe. Ohio State was deserving based on their victory over Miami, but there is little reason to believe they would fare well against any other champion of the BCS era.

Given their track record over the first hundred plus years of Buckeye football, I would particularly warn them to steer clear of the 2003, 2006, 2007, and 2008 champs.

Michael said...

1. Yes, Miami had won 34 in a row going into the '02 title game, but Butch Davis had been their coach for 25 of those wins. The '02 Miami team was vulnerable, as evidenced by the fact that they trailed Rutgers for a long stretch and barely beat a mediocre VT team. The same can be said for '05 USC. That defense was so bad that a quality team stood a good chance of winning a shootout against them.

2. I'd be interested to know what a rational Ohio State fan would say about the following: Ohio State being lucky in 2002 convinced Jim Tressel that he could win based on a solid defense and special teams, which then held OSU back for years thereafter. I'm a Michigan grad and I would absolutely make the case that the '97 national title was bad for Lloyd Carr because he took from it the lesson that he could win with a bland Mike Debord offense. As a result, Michigan teams with great offensive talent and less defensive ability ('99, '00, and '07) all underachieved.

TrevJo said...

So, where's the 1994 Michigan data?
What about the 90s national champions?

LJ said...

Nice analysis Michael. One thing this doesn't take into account is special teams performance. Have you considered looking at average yards gained/lost per special teams possession change?

Josh said...

"Josh, wins are a statistic, just like points and yards. They are also a stat containing a tremendous amount of noise. There has been a ton of research by baseball statheads and there is overwhelming evidence that run differentials are a better measure of a team's future performance than record. Digging deeper, underlying stats like offensive and defensive performance can be a better measure than runs. Football is no different.

Right, playing the Citadel is SO different than playing Youngstown State. Are you really going to take the position that Florida has far superior offensive numbers than Ohio State because of the plays they call for their scrubs in blowouts? If you really think that, then we have nothing to discuss because you have your head firmly planted in the sand. Or maybe you missed the the last Ohio State-Florida game."


#1 2002 Ohio State DIDN'T PLAY Youngstown State, there were no Division 1-AA opponents on our schedule in 2002. Who our teams scheduled in the future had nothing to do with the statistics you collected

#2 You have too small of a sample size with statistics like this your looking at only 13 or 14 games played and theres too much a variance between opponents to draw a decent conclusion. This isn't baseball the 2 sports are different. As mentioned by someone else offensive and defensive performance are not 2 seperate things, when an offense isn't as good a defense is going to tend to give up more yards/points (always being on the field and getting tired etc)

#3 Again, as someone mentioned, special teams were left out of the discussion, special teams are a gigantic part of games. If you can't play special teams you aren't going to win shit.

#4 The 06 Ohio State-Florida game has nothing to do with how good Ohio State was in 02. 06 Florida was probably better offensively then the 02 Ohio State team, but defensively OSU was probably better and were definately better on special teams.



#5 If these stats are supposidly so accurate at predicting success then 07 LSU wouldn't have lost those 2 games, 02 OSU would have lost games and 08 Florida wouldn't have gotten beat by Ole Miss

LJ said...

Josh--I disagree, I think the analysis is very valuable. Here's why.

1) Regardless of whether they played a 1-AA, OSU had some non-conference cupcakes on the schedule. They beat Kent State 51-17 and San Jose State 50-7. That is beating up on easy competition just like beating up on the Citadel is. I agree with Michael that the differences in non-con scheduling is not driving the statistical differences.

2) Here, you are actually making Micael's point. As he said, using the statistic of points scored in a game is very noisy with a sample size of only 13-14 games. However, using Michael's yards per play analysis, each play is a sample, making the sample size hundreds of times larger, and increasing the reliability of the statistic. Because there are so many plays, flukes and outliers are weighted less significantly.

3) Special teams are a concern--I made that comment earlier. However, I doubt that the difference is enough to make up for OSU huge yard differential gap between them and the top group of championship teams

4) Agreed--that game is irrelevant when discussing the '02 OSU team. I think Michael was just trying to make a point.

5) Again, this goes back to the point that wins and losses are noisy and greatly affected by chance. OSU won many close games in 2002, and LSU lost a couple. The model can't predict wins and losses perfectly, but the model would predict that if 2002 OSU and 2007 LSU both played 1000 games, LSU would come out with more wins, and I would certainly bet on LSU in that situation.

Michael said...

I second what LJ said. The experience of being a Braves fan in the 90s made me a fan of looking at the larger sample size rather than the smaller one. Looking at plays is a bigger sample size than looking at record. Josh is exactly right that college football has a small sample size of games, which supports looking at the larger sample size of plays as opposed to the smaller sample size of wins and losses. I'm not saying that the better team always wins or that we shouldn't look at record when determining a national champion. I am saying that when we compare champions across a decade, looking at the strength of those teams by evaluating per play differentials is a good measuring stick.

Michael said...

And yes, it would be better if I could measure special teams, as well. That is a valid criticism. I just don't know of a good measuring stick for special teams that would take into account the various elements.

Unknown said...

Washington needs to be on that list.

Louise said...

As a baseball fan, I have the opportunity to learn about Braves my best team , including getting the latest news, schedules and also watching their hottest videos .all this is within a touch in an interesting website I lately discovered:
http://www.bravesworld.com

hugh said...

Lollin' at the dopes who don't understand statistics coming out of the woodwork here.

Michael said...

As someone who dropped Stats 402 at Michigan after a week, I hope that I'm not one of the dopes. :)

Burke said...

I think this is a good analysis.

If you want to improve it, as someone wrote, show the hidden yardage stats for special teams. Then crank in something for Red Zone offense and defense play, field goal percentage, and turnover ratio.

Many coaches would regard the turnover ratio as the most important factor of all.

Hobnail_Boot said...

Of course the white elephant in the room is that Ohio State got the most gratuitous/questionable penalty in modern college football history to get that crystal football.

They are the exception that proves the rule, in every way.

Michael said...

I'll actually defend OSU on the call in overtime of the national title game. It was a bad call, but OSU got hosed on a pass completion to Michael Jenkins at the end of regulation that would have allowed them to run out the clock.