Monday, June 18, 2007

Diagnosing Turnover Margin

Statistically Speaking recently examined fumble luck--how do teams that are lucky/unlucky in terms of fumble recovery percentage fare the next season? Not surprisingly, he finds that teams generally regress back to the mean (i.e., to a recovery percentage around 50%). The list of the unluckiest fumble recoverers in 2004:

2004 Worst Fumble Recoverers
Team/Fumble Recovery %/2004 Record

Ohio State/34.3%/8-4
NC State/35.1%/5-6
Georgia Tech/35.7%/7-5

This led me to dig a little deeper. As you undoubtedly know, since Philip Rivers graduated, we've made a habit of digging ourselves a hole with turnovers. Not only have we had problems with giveaways, we've also been terrible at forcing opponents to make mistakes.
       TO_Margin (Rk)
2000 0.36 (39)
2001 0.91 (13)
2002 0.5 (35)
2003 0.15 (53)
2004 -1.55 (114)
2005 0.0 (55)
2006 -0.92 (112)
So what's the deal? There are several factors at work, and I'll start in the obvious place.

1.) The Jaycus Evans Effect

In Philip's four years at NC State, the Pack threw a total of 36 interceptions (fn. 1)--an average of less than 10 per season. From 2004 onward, it's been a bit of a different story.
        G    INTs  PassAtt  INT/100PA
2000 11 11 455 2.4
2001 11 7 385 1.8
2002 14 11 447 2.5
2003 13 7 496 1.4
2004 11 16 343 4.7
2005 12 14 335 4.2
2006 12 16 379 4.2
The noteworthy column is the last one, interceptions per 100 pass attempts. It's pretty easy to tell when a certain someone graduated, isn't it? Rivers averaged right around 2 INTs per 100 attempts; the Davis/Stone/Evans mishmash has thrown INTs twice as often. Not only are we getting less value from those guys in terms of yardage per attempt and completion percentage, we're also having to deal with more frequent giveaways.

2.) Lady Luck Says We Can Go %#$@ Ourselves

Staying on the offensive side of the ball, let's shift to fumbles. It turns out that one of the reasons those Rivers teams were so good at hanging on to the ball was simply because they were fortunate.
        G    FumL FumTot  Rec%    Ru    Fum/100Ru
2000 11 9 25 64.0 351 7.1
2001 11 6 22 72.7 355 6.2
2002 14 13 27 51.9 522 5.2
2003 13 12 27 55.6 392 6.9
2004 11 16 23 30.4 435 5.3
2005 12 10 19 47.4 440 4.3
2006 12 7 15 53.3 374 4.0
In Rivers's freshman season, State fumbled 25 times, losing a mere nine of them. We were even luckier the next season, recovering a staggering 72.7% of our own fumbles. Note that recent Wolfpack teams have been far better at hanging onto the ball than the Rivers-era teams (see fumbles per 100 rush attempts); the problem is that the football gods have deigned to smite us. The 2002 and 2004 offenses fumbled at the same rate, yet, because of the way the cookie crumbled, the '04 Pack lost three more fumbles (in four fewer opportunities) than the '02 Wolfpack.

3.) A Lesson In Stripping

There are, of course, two sides to the turnover margin coin: the offense's giveaways and the defense's takeaways. We've been blessed with some very good defense over the last few years, but--and this is the head-scratcher--these defenses haven't been able to force mistakes for the life of them. More specifically, these defenses haven't been able to force fumbles.
     Opp_INT/100PA  Opp_Fum/100Ru
2000 3.1 5.5
2001 2.0 4.4
2002 3.7 5.9
2003 2.1 5.7
2004 3.3 3.3
2005 3.4 3.7
2006 2.6 2.6
The rate at which we've intercepted the ball has essentially been constant over the last seven years, so that's not the cause of our post-Rivers woes. After the 2003 season, the rate at which we've forced fumbles has fallen off a cliff. Take a look at the breakdown of fumbles forced by season:
2000     28        
2001 20
2002 31
2003 27
2004 14
2005 17
2006 13
It's not that the post-Rivers defenses have been unlucky in recovering fumbles, it's that the opportunity to recover them hasn't been there. The defensive recovery percentages:
Rivers Era     52.0
Post-Rivers 50.0
We forced more than two fumbles per game in three of Philip's four seasons; since then, our best mark is the 1.4 FPG we forced in 2005. While the quarterbacks have been busy throwing INTs twice as often, the defense has been causing fumbles half as often.

Summing Up
The offense saw its interception rate double after Philip Rivers departed and lost its fumble recovery mojo at the same time, while the defense maintained a consistent interception rate but had the damnedest time making opponents put the ball on the carpet. Put those things together and you've got the horribly negative turnover margin we've been fighting for three seasons now.

A few notes:

-- The luckiest fumble-recovering team of those I examined was easily the 2001 Wolfpack. In addition to the offense's 72.7% recovery percentage, the defense somehow managed an 80% (16 of 20!) recovery percentage. All told, that team recovered 32 of 42 fumbles (76%). I can't help but think of that year's Florida State game--as Cotra Jackson neared the end zone in the first quarter, he was stripped, and the ball rolled over the goalline. Miraculously, a State offensive lineman (Derek Green?) fell on the ball, scoring the Pack a TD. Definitely a microcosm of the season, that play.

-- It's harder than I would have expected to find evidence of TA McLendon's (negative) impact. I mean, it's easy enough to see that replacing him with the sure-handed Andre Brown made a difference in our fumble rate, but in looking at just the seasons during which TA was on the team, our highest fumble rate came in the year he missed the most playing time.

-- That said, I should point out that the rush attempt totals include rushes by all players, not just running backs. Quarterbacks therefore have a small effect on fumble rate, and this perhaps makes our 2005 and 2006 numbers a little more impressive (I'm assuming here that Rivers fumbled less often than Davis/Stone/Evans, which may not be the case). If the NCAA's database listed fumbles for each player, I could speak with greater certainty, but alas, it does not.

-- If you've got any theories as to why the defense has seen such a drop off in forced fumbles, I'd love to hear 'em.


1.) The NCAA's database doesn't include the Pack's bowl games in 2000 and 2001, so those games are not factored into the averages for those seasons.