This week's DVOA commentary is all about worsts. Come find out where Washington stands among the worst special teams in DVOA history, whether San Diego has the biggest gap between offense and defense, and whether Baltimore or Jacksonville has the worst running game we've ever tracked.
25 Aug 2009
By Bill Connelly
It is the easiest prediction in the world to make: Team A looked great in its bowl game, or it got hot over the last handful of games, so therefore the team is going to do big things this year. The team struggled early in the season, but the players and staff righted the ship and turned the corner, and now there's no turning back.
Or, at the other end of the spectrum, Team B started fast but faded and faces a new year with no momentum. The team was exposed during its late-season struggles, and things aren’t looking good in the new season.
Is it an easy prediction because it's grounded in fact. Or is it just lazy? Is it possible to have momentum (good or bad) to start the season when you haven't played a game in eight months?
Does end-of-year momentum affect the next year's performance?
To try answering that question, we're going to look at all teams' S&P+ rankings only for games that took place between Nov. 1, 2007 and the end of the 2007 bowl season. We’ll look at that team's overall 2007 S&P+ performance, then compare that to how it ended up doing in 2008.
|Late-2007 S&P+ rankings and their impact on 2008|
All of 2007
So being one of the top teams at the end of the season makes a difference (USC, TCU) ... except for when it doesn't (East Carolina, BYU). That doesn't really tell us a lot. In all, here's a look at some 2007 vs .2008 correlations:
What about offensive and defensive breakouts? Does late-season success have an impact on one side of the ball more than the other?
In all, there is at least a decent correlation between late-season success and success the following season, but looking at the whole season's work tells more of a story.
Let's look at this a different way, sorting teams by biggest late-season improvement.
|Biggest Late-2007 Improvement (+5% or more), BCS conference teams*|
Change in S&P+
|* For Varsity Numbers columns, "BCS Conference teams" is used
rather loosely to include both BCS teams, Mountain West teams,
and notable non-BCS teams like, in this case, East Carolina.
Overall, Florida International improved the most at the end of 2007 (+18.3 percent), and that success continued into 2008 (+14.0 percent). Including non-BCS teams, 29 teams improved by at least 5 percent at the end of 2007. Eighteen of them improved overall in 2008, seven by more than 10 percent. On the flipside, five teams regressed by at least 5 percent in 2008 (North Texas was the worst, at -8.5 percent).
Of the eight teams that improved by at least 13 percent at the end of 2007 (Florida International, Tulane, Western Michigan, Eastern Michigan, Army, Memphis, Marshall, Utah), seven improved in 2008, and by an average of 5 percent.
(It should also be noted that East Carolina started the 2008 season as hot as hot can be, beating Virginia Tech and destroying West Virginia before losing three in a row and finishing 9-5.)
So now let's look at the opposite end of the spectrum.
|Biggest Late-2007 Regression (-10% or more), BCS conference teams|
Change in S&P+
Dealing with late-season regression is a completely different animal, for three main reasons:
1. The changes are bigger and more frequent (41 teams regressed by at least 5 percent).
2. The slip could be caused by injuries (see: Oregon losing quarterback Dennis Dixon to a leg injury; North Carolina’s quarterback T.J. Yates fighting through a shoulder injury; Auburn's Brandon Cox separating his shoulder).
3. Major regression could get a coach fired (see: Baylor), bringing major turnover from year to year and making the resulting change a little less relevant.
That said, it is certainly notable that five of the six BCS conference teams that regressed at least 13 percent at the end of 2007 (and six of seven overall, including Kent State, which regressed 20.4 percent), saw further regression in 2008. There are a lot of ups and downs in the data, but it seems that if a team improved or regressed by at least 13 percent at the end of 2007, it was probably going to continue that upward or downward movement in 2008. Call it The Rule of 13%. With more data, we might see a different trend, but for now this is the reality.
This begs a pretty obvious question: who improved or regressed the most at the end of 2008?
|Biggest End-of-2008 Improvement, BCS-conference teams
Wyoming (the only team to which The Rule of 13% apparently applies) and Auburn changed coaches in the offseason, Duke and Utah lost a majority of their playmakers, and West Virginia lost Pat White. Clearly this data suggests more for some teams than others. It does say very good things about Stanford's prospects for improvement in 2008, not to mention N.C. State, Florida State, and Notre Dame.
Biggest End-of-season Offensive Improvement
1. TCU (+24.1%)
2. Florida State (+15.1%)
3. Notre Dame (+14.6%)
4. Stanford (+13.5%)
5. Auburn (+13.4%)
TCU was carried by its rather ridiculous defense, but the offense certainly began to figure things out late in the season, putting up 44 points against UNLV and Air Force after Nov. 1. Meanwhile, Florida State, Notre Dame, and Stanford all began to click. FSU has to answer questions at the skill positions (the team lost its top two receivers and its leading rusher), but the improvement of quarterback Christian Ponder and an extremely experienced offensive line should keep the chains moving for the Seminoles.
Meanwhile, we all heard about Lou Holtz's certifiably insane national title prediction. Though a national title may not be in the works for Notre Dame, nobody is doubting that the team should continue to improve in 2009, along with the aforementioned Stanford Cardinal.
Biggest End-of-Season Defensive Improvement
1. Wyoming (+23.6%)
2. West Virginia (+19.4%)
3. N.C. State (+16.1%)
4. South Florida (+14.9%)
5. BYU (+11.9%)
Former Missouri offensive coordinator Dave Christensen will try to bring offensive firepower to Wyoming. If he wrings some yards out of the offense, it will help to balance a defense that was hot at the end of 2008 and returns eight starters. Meanwhile, West Virginia's improvement sprung mostly from a miserable defensive start to the season (they improved to average), but N.C. State returns some key cogs to a hot late-season defense, including linebacker Nate Irving and defensive end Willie Young.
|Biggest End-of-2008 Regression, BCS-conference teams
It's hard to ignore the presence of multiple Big 12 teams here. Texas and Missouri cooled off in November -- including a couple of pretty poor offensive efforts in bowls versus Big Ten teams -- after setting an unsustainable early pace on offense, while Kansas State all but quit on Ron Prince. Iowa State and Colorado were two bad teams that simply got worse.
As for the other conferences, Arkansas finished the season poorly, with losses to South Carolina and Mississippi State, while Arizona State made the list for scoring only 39 and 31 against Washington and Washington State, respectively, then laying a complete egg against Arizona.
Biggest End-of-Season Offensive Regression
1. Michigan State (-28.0%)
2. Kansas State (-26.8%)
3. Missouri (-21.6%)
4. Texas (-21.3%)
5. LSU (-17.8%)
It is certainly worth noting here that Texas has now come up in two red-flag categories -- significant end-of-season regression and disproportionate success on Passing Downs. We will see how much of an impact this makes on the Longhorns' title chances -- their schedule is light on landmines (toughest out-of-conference game: a trip to Wyoming), but all indications point to regression on offense. If they are lucky, or if their defense turns out to be really good, it probably will not matter.
It is hard to draw too many other conclusions among these five teams. Michigan State relied so much on Javon Ringer that his departure will result in a completely different offensive mindset. Meanwhile, Missouri is replacing most of the players that were so good in September and parts of October, and so iffy in November and December.
LSU's presence here is surprising when you think about how great they looked in crushing Georgia Tech in the Chick-Fil-A Bowl, but they were pretty miserable in November, losing three of five games.
Biggest End-of-Season Defensive Regression
1. Ohio State (-24.6%)
2. Purdue (-24.2%)
3. Kentucky (-23.5%)
4. Iowa State (-23.1%)
5. Washington (-22.3%)
Kentucky will have a potentially strong defensive line in 2009, not to mention preseason All-SEC corner Trevard Lindley, but they were quite unimpressive down the stretch, giving up an unforgivable 59 combined points to Vanderbilt and Tennessee. But they did close the season out with a nice bowl performance in the Liberty Bowl against East Carolina. Meanwhile, don't look too closely at the top two teams -- like Missouri on offense, Ohio State's defense is undergoing enough turnover that the poor performance down the stretch won't mean a lot. (Then again, we are talking about some pretty serious regression here.)
The other three teams on this list -- Purdue, Iowa State, and Washington -- went from bad to putrid, so we will see what that means for 2009.
Within the next six months, not only will we have 2009 play-by-play data to play with, but we will also have 2005 and 2006 data completed. With five years of data, we can look closer at potential red flags mentioned here and elsewhere: things like momentum, returning starters and disproportionate success on Passing Downs, etc. But even with only one season-to-season transition to study, a certain narrative does still emerge. You cannot necessarily look at records or bowl performances to gauge momentum from one season to another, but you can likely judge that teams that improved or regressed significantly in terms of S&P+ (for now, the baseline is set at around +/- 13 percent) will pull some momentum, good or bad, from that. Still, most teams are much more correctly judged by how they performed in the whole season, not just during the cold-weather portion of the schedule.
5 comments, Last at 05 Sep 2009, 3:15pm by firstname.lastname@example.org