Phil Steele's 2016 "Turnovers=Turnaround" article: Michigan set to improve

Submitted by markusr2007 on July 16th, 2015 at 1:55 PM

Phil Steele published his popular "Turnovers = Turnaround" piece today.

Link: http://www.philsteele.com/Blogs/2015/JULY15/DBJuly16.html

One of the most popular articles that appears annually in my magazine is entitled “Turnovers=Turnaround.”  The basic premise of the article is simple:  teams that benefit from a large turnover margin in a particular year are not likely to enjoy that same margin in the upcoming year.  As a result, a team’s record is more likely to decline if it relied heavily upon turnovers as a core basis for winning.  Conversely, a team that struggles with an unfavorable turnover margin will often experience a reversal of fortunes.  Since turnover margin is such a critical factor in a football game, a change in turnover margin can lead to dramatically different results.

This is not to say that turnovers are a complete fluke.  In fact, certain teams have been able to create turnovers somewhat consistently on a year-to-year basis, usually via interceptions as opposed to fumbles.  Again, fumbles are not completely random and some teams are better than others at stripping (or protecting) the ball, but it is a bit of a toss-up as to whether a team will be able to recover a fumble once an oblong ball starts bouncing around on the ground.

The examples are limitless but the core theory is simple and has proven to be accurate year-in and year-out.

Going Down?
Oregon +23
Michigan State +19
TCU +18
Georgia +16
Louisiana Tech +16
Arizona State +14
Baylor +13
Northern Illinois +12
Washington +12
FIU +11
Georgia Tech +11
Memphis +11
Nevada +11
USC +11

Going Up?
Georgia State -22
Eastern Michigan -18
Washington State -17
Michigan -16
Vanderbilt -16
West Virginia -15
SMU -14
Connecticut -13
New Mexico State -13
Texas Tech -13
San Jose State -12
Idaho -11
Colorado -10
Southern Mississippi -10

Comments

MayOhioEatTurds

July 16th, 2015 at 2:00 PM ^

season:  Harbaugh coaching quarterbacks alone ought to help.  (And that improvement ought to help with the record, too.) 

Anyone want to take the bet?  House for house?

justingoblue

July 16th, 2015 at 2:33 PM ^

I'm sure there are diaries and stories from this site that illustrate the same general point, but turnover margin and wins are pretty strongly correlated.

http://www.footballstudyhall.com/2013/8/23/4649718/college-football-tur…

Not all of that has to do with turnovers in a vacuum (having Charles Woodson means you get interceptions + you've got Charles freaking Woodson on every non-interception play) but I'd bet in many or even most years there would be two of the top four teams with a top ten turnover margin.

LSAClassOf2000

July 16th, 2015 at 3:04 PM ^

If you calculate R^2 for our own average turnover margin compared to win totals over the last ten seasons or so (I went back to 2004 anyway), it comes out to 0.7898, which is not bad at all considering the confounding things that happen in football games and therefore get mixed into such comparisons in the first place. It's a pretty similar figure throughout the Big Ten from what I recall anyway, although in different confierences it might vary more. 

alum96

July 16th, 2015 at 4:05 PM ^

But a lot of top 10ish teams did make it - Baylor (7), MSU (5), AZ State (12), GA Tech (8), Georgia (9), TCU (3) along with Oregon (2).

Put another way of the top 14 of 128 teams in college football, 6 were in top 9.  That's definitely correlation.  As you stated the other 3 were in the playoffs - a weird quirk.

evenyoubrutus

July 16th, 2015 at 2:04 PM ^

This is good stuff but I would be even more fascinated by a study that shows previous years turnover statistics and how common it is for teams to regress back to the mean and whether we SHOULD expect improvement.

NittanyFan

July 16th, 2015 at 2:39 PM ^

the correlation coefficient between turnover margin (per game) in 2013 and turnover margin (per game) in 2014 was 0.296.  (edited because originally I had R^2).  This number is signficiantly higher vs. the NFL number.

The linear-regression equation:

intercept = 0.00439, coefficient on last year's turnover margin variable = 0.31364. 

t-stat on the last year's turnover margin variable = 3.44.  Which is actually fairly high.  

Predicted turnover margin (per game) for Michigan in 2015, based on that equation and the U-M numbers in 2014 = -0.41275.

Source of the stats is here (throw out App State, GA Southern & ODU as they were not FBS in 2013):

http://www.cfbstats.com/2014/leader/national/team/offense/split01/categ…

http://www.cfbstats.com/2013/leader/national/team/offense/split01/categ…

Blue_sophie

July 16th, 2015 at 2:10 PM ^

Seems like regression to the mean arguments should focus on something like fumble recoveries, where the outcome is much more random. Of course you would need to also include fumbles that Michigan recovered. Same thing could be done for jump balls. I doubt that data is all in one place, however. Not sure how far analytics and video tracking has progressed for college football. Would be interesting to crowd-source that type of research.

DavidP814

July 16th, 2015 at 2:32 PM ^

It should be.  I used to love Phil Steele's magazine about 10 years ago, but most of the statistics he cites are extremely elementary and has been surpassed many times over by the guys at FootballOutsiders and others who use high-level analytics as a basis for their predictions.

Everyone Murders

July 16th, 2015 at 2:11 PM ^

I think that turnovers are somewhat random, but not entirely so - which is in keeping with Steele's premise. 

Much of it depends on personnel.  If Mike Hart is your principal running back, your turnovers are going to decline.  Full stop.  If [insert QB from last 7 years] is your QB, you're going to give up an incremental number of picks.  These are not fully random events, although there is some randomness involved.

In any event, I expect that Harbaugh is not above quoting John Heisman:  "Gentlemen, it is better to have died as a small boy than to fumble this football."  -16 should improve.

 

Pelini's Cat

July 16th, 2015 at 2:15 PM ^

Ok but I think that when you have a defense as good as state's, or QB play as bad as ours, interceptions are less of a fluke. That said, we will hopefully put our QBs in less situations where they are forced to make bad throws

Sent from MGoBlog HD for iPhone & iPad

Hannibal.

July 16th, 2015 at 2:19 PM ^

I would expect us to regress to at least zero for a few reasons:

1.  Better QB coaching and a fifth year senior QB who had a 3:1 TD:INT ratio at Iowa

2.  We were -3 on fumbles

3.  Not only was our QB horribly turnover prone, but our defense was terrible at forcing them.  We only picked off five passes last year.  Two of them were by defensive linemen.  Only one DB picked off a pass.  That's unbelievaby bad.  I'll bet that even in the 1970s when QBs only through the ball 10 times a game we never had a season where only one DB intercepted a pass. 

WolverineHistorian

July 16th, 2015 at 2:43 PM ^

All I know is, I'm burned out on turnovers.

During the last 5 seasons, fumbles lost have been 8, 8, 7, 5 and 14.

Interceptions have been 39 by Denard, 32 by Gardner, 5 by Morris and 4 by Forcier.

That equals to 122 turnovers in 5 seasons. I have no idea how that compares to the rest of college football but I doubt many teams rank above us in this category during this time frame.

Sent from MGoBlog HD for iPhone & iPad

alum96

July 16th, 2015 at 4:25 PM ^

I have to imagine outside of a team such as Kansas or similar ilk which win 3-4 games a year we have been the most turnover program in P5 in the past 7 years

Last year was worse than normal as both units sucked at turnovers (other years at least the D offset the very bad TO margin from the O to a degree) and you had last year.  If we had just a "national average" turnover rate I imagine we'd have 3 more wins last year alone.  I guess perversely that would not have been a good thing in the long run based on Hackett's affinity for bringing back Hoke if he had done just a tad better.

  • 2014: -16  (121st in nation)
  • 2013: +5  (32nd in nation)  [defense created 26 TOs to offset offense's 21]
  • 2012: -9 (99th in nation)
  • 2011: +7 (25th in nation)
  • 2010: -10 (109th in nation)
  • 2009: -12 (115th in nation)
  • 2008: -10 (104th in nation)

Except for 2 outliers we have been 99th or worse (!!!) in 5 of the past 7 yrs.

wahooverine

July 16th, 2015 at 2:53 PM ^

The offensive side of the equation should pull the margin closer to even. The defense had 5 picks and 5 recovered fumbles.  The offense threw 18 Ints last year (Gardener-15, Morris-3) and had 15 fumbles (8 lost) for a total of 26 takeways. Many of the offenses turnovers were game changing, momentum killing plays from the QB, especially ludicrous fumbles during scrambles.  Those are just the facts from last year and I've done no statistical analysis to uncover how flukey or not flukey these numbesr are. Yet from what I saw, I think it's reasonable to think expect a turnover margin closer to even. The QB should be in better position to make easier reads and better throws; he should get better protection (and not be shellshocked) and should benefit from a better run game. If it's Ruddock, there is reason to believe he is accurate and protects the ball. I'd expect us to more consistently be ahead of the chains and not playing from behind.  Ty Wheatly will coach his backs to protect the rock. Assuming all this happens who knows how it translates into W&L's with a tougher slate, however I'd be shocked with anything less than 7 wins and wouldn't be surprised by 9 or 10.

jaydubya

July 16th, 2015 at 3:46 PM ^

Reducing offensive turnovers (note the double meaning) is the biggest reason why Rudock must be the starting QB this year, in my opinion.

ElBictors

July 16th, 2015 at 4:11 PM ^

Reminds me of a Multiple Regression Analysis group project in Grad School where we took a decade of NFL stats in order to determine the key statistic in terms of predicting SuperBowl winning success ...

After all the data crunching and pain of the data input to begin with, the data concluded it was RUNNING YDS/GAME.

In other words, for SuperBowl teams up through about 2004, the key factor was a running game...    Duh..

 

To the OP and the article, I do think programs like Sparty have become good at forcing turnovers, akin to a VT sort of history.  I predict M's Defense will be a lot more disruptive and force the game more and Ruddock will simply not repeat Gardner. 

Has to come down from -16

#MeanReversion

Red is Blue

July 16th, 2015 at 5:06 PM ^

Turnovers can also be an effect of winning and losing, not just a cause.  If you are losing and time is becoming a factor, you might start calling riskier plays and embrace a higher likelihood of turning the ball over.  Conversly, if you're winning, you might call more conservative plays with a lower probability of turning the ball over.

Avon Barksdale

July 17th, 2015 at 2:29 PM ^

Because the biggest influence in going from 7-6 to 11-2 in 2010 to 2011 was the cut down of offensive turnovers vs finally forcing turnovers on defense. The defense in 2011 wasn't elite, but they forced enough turnovers to get Michigan to the Sugar Bowl.

We went from 109 in turnover margin to 34 basically overnight.