Compu-Picks Analysis: Turnovers

Mr Pac Ten
Posted May 15, 2011


2011 Compu-Picks Analysis Part 3: Why Turnovers are a Valuable Part of Pre-season Predictions

Any of you who regularly read Phil Steele's annual preview magazine (or regularly visit his website and saw the 2009 or 2010 articles) probably noticed that one key point of emphasis is turnover margin. Specifically, Phil Steele suggests that teams who enjoy an especially good turnover margin in one year tend to have a worse record the next, because the turnover margin tends to decline (and the opposite for those who suffer an especially bad turnover margin).

As part of creating a preseason compu-picks model, I looked at a number of potentially relevant predictive variables. Among these variables was turnover margin. Was Phil Steele right? Does turnover margin really have a legitimate predictive value for the next year?

The answer is yes. Turnover margin has meaningful predictive value. The reason for this is twofold: first, that turnover improvement very strongly correlates to record improvement; and second, that the turnover margin from one year generally has a very small correlation with the turnover margin for the next year. Let's look at those two points in more detail:

1) Turnover improvement very strongly correlates to record improvement.
As a test of this, I took at look at the ten teams who improved their turnover margin the most between 2009 and 2010, and looked at how their W/L records changed in the same time frame. They were as follows:

MiamiOH
2009: 1-11 (-24)
2010: 10-4 (+11)

Georgia
2009: 8-5 (-16)
2010: 6-7 (+10)

Hawaii
2009: 6-7 (-11)
2010: 10-4 (+12)

Maryland
2009: 2-10 (-6)
2010: 9-4 (+15)

Toledo
2009: 5-7 (-10)
2010: 8-5 (+11)

Tulsa:
2009: 5-7 (-1)
2010: 10-3 (+17)

NC St
2009: 5-7 (-11)
2010: 9-4 (+7)

WKU
2009: 0-12 (-12)
2010: 2-10 (+3)

Tulane
2009: 3-9 (-16)
2010: 4-8 (-2)

Stanford
2009: 8-5 (+0)
2010: 12-1 (+13)

Nine of the ten improved their record. Six of the ten improved by 4+ wins. And in case you're wondering, it was a similar story with those who improved by 12 turnovers (OK St, Illinois, and Mizzou all materially improved their records as well).

And on the other side, here at the nine worst drops in turnover margin:

MTSU
2009: 10-3 (+12)
2010: 6-7 (-19)

Cincy
2009: 12-1 (+9)
2010: 4-8 (-15)

Texas
2009: 13-1 (+9)
2010: 5-7 (-12)

Ohio
2009: 9-5 (+13)
2010: 8-5 (-7)

ECU
2009: 9-5 (+11)
2010: 6-7 (-7)

UCLA:
2009: 7-6 (+5)
2010: 4-8 (-11)

Air Force
2009: 8-5 (+22)
2010: 9-4 (+5)

CMU
2009: 12-2 (+5)
2010: 3-9 (-11)

UAB
2009: 5-7 (+9)
2010: 4-8 (-6)

This is a similar story to the top ten improvers, only in reverse. Eight of the nine worsened their record. Four of the nine got worse by 4+ wins, and another two got 3 wins worse. Of those whose turnover margins dropped by 13 or 14, Georgia Tech and Rutgers got a lot worse, SMU got a bit worse, Boise hovered, Indiana got a bit better, and Arkansas improved materially.

Overall, there's a very clear pattern to these results. Teams that get a lot better in turnover margin tend to get a lot better in record. And teams that get a lot worse in turnover margin tend to get a lot worse in record. And this leads us to the second important point.

2) The turnover margin from one year generally has a very small correlation with the turnover margin for the next year.

As a test of this, I decided to run a correlation between one year's turnover margin and the next year's. As shown in the table at the bottom of the article, I had eight years of turnover margin data for 120 teams. Since we don't know 2011's numbers, I couldn't use 2010 as a baseline, leaving me with 7*120 = 840 independent data sets, plenty to run a simple correlation model.

So running the correlation between one year's turnover margin and the next, I found that the correlation was a mere 12%. That's still something, but it's clear that for most teams, the turnover margin they enjoy one year has virtually zero predictive value for the turnover margin they will enjoy the next year. That means that on average, teams with substantially positive margins will see major decline in margin the next year, and teams with substantially negative margins will see major improvement the next year. A team with a -10 turnover margin in 2009, for example, would have an expected turnover margin of -1.2 in 2010, an improvement of nearly a full turnover per game!

And, in fact, there were three such teams in 2009: Toledo (+11 in 2010), San Diego St (-6), and Fresno St (-11). Their collective average was -2.0, very close to the predicted -1.2.
In 2008 there were three such teams: Michigan (-12 in 2009), Miami (0), UCLA (+6). Again, a -2.0 average.
In 2007 there were three such teams: Ole Miss (-2 in 2008), Army (-14), Marshall (-4). This time, a -6.6 average.
In 2006 there were two such teams: Oregon (+9 in 2007), New Mexico St (-15). This time, a -3.0 average.
In 2005 there was only one such team: UTEP (0 in 2006).
In 2004 there were four such teams: Michigan St (-1 in 2005), MiamiOH (+13 in 2005), BYU (+2 in 2005), USF (+5 in 2005). This time, a substantially positive average, +4.75.
And in 2003 there was only one such team: Vandy (+4 in 2004).

Overall, you get a next year turnover margin that is on average just a bit less than zero (I got -0.88 averaging by team). Some substantially negative, some substantially positive, and many either zero or very close to it. Of course, that's just one set of examples, but it nevertheless demonstrates fairly clearly how little value turnovers in one year have in terms of predicting turnovers the next.

All that said, there do appear to be some exceptions, teams that have enjoyed substantial long-term success (or at least if you define 8 years as long-term), as well as those who have suffered substantial long-term failure in turnover margin. It is difficult to accurately assess just how much of the success was noise and how much of it was legitimate and likely to continue going forward, but at least some of it looks real.

That said, unless you're in the top ten (MAYBE fifteen) in terms of long-term turnover average, it seems fairly safe to say that if you enjoyed a double-digit turnover margin in 2010, you're likely to suffer a substantial decline in turnover margin in 2011. Ohio St (+15), Mizzou (+11), Iowa (+13), NIU (+11), Oregon (+13), UConn (+12), Georgia (+10), Wisconsin (+14), Oklahoma St (+12), Stanford (+13), Toledo (+11), Tulsa (+17!), Maryland (+15), Hawaii (+12), and especially Army (+16, with an eight-year average of -5), be warned: the turnover success you enjoyed last year is unlikely to happen again. And if it doesn't, then you'd better hope you're like 2010 Air Force instead of 2010 Texas or Cincy.

And for those of you who like to play along at home, here is the data I used (source: Phil Steele mags for 2003-2009 [there may be an occassional typo], cfbstats.com for 2010).

Turnover Margin - by team by year

TEAM NAME Average 2010 2009 2008 2007 2006 2005 2004 2003
VIRGINIA TECH 9.8 19 9 14 11 4 9 13 -1
USC 9.6 4 0 7 2 4 21 19 20
FLORIDA 8.8 2 7 22 5 5 18 4 7
OKLAHOMA 8.5 14 4 23 8 -1 -1 4 17
ALABAMA 7.8 11 19 6 4 7 8 6 1
SOUTHERN MISS 7.6 9 10 13 -1 6 14 5 5
BOISE ST 7.5 8 21 7 1 11 -8 10 10
UTAH 7.4 -1 5 13 11 8 -1 15 9
AIR FORCE 7.3 5 22 13 10 8 -7 1 6
WEST VIRGINIA 7.3 -5 -2 12 13 7 14 3 16
TCU 6.6 9 2 13 -7 7 21 4 4
WAKE FOREST 5.9 0 -4 17 9 13 -2 7 7
OHIO STATE 5.6 15 17 16 -3 9 -9 -1 1
MISSOURI 5.1 11 4 -4 13 2 -1 5 11
IOWA 4.5 13 2 8 8 -11 -1 13 4
NAVY 4.4 7 9 15 -2 2 -6 3 7
BOSTON COLLEGE 3.5 8 -3 3 6 15 -4 0 3
N ILLINOIS 3.3 11 7 2 -17 2 -1 8 14
LSU 3.1 8 4 -1 20 0 -9 -2 5
NOTRE DAME 3.1 1 5 -3 0 5 10 5 2
OREGON 3.1 13 2 5 9 -10 13 -2 -5
PITTSBURGH 3.1 1 10 -3 -5 8 -4 13 5
CALIFORNIA 2.9 2 4 15 -3 6 -6 2 3
CONNECTICUT 2.9 12 3 -4 14 -1 4 -5 0
SAN JOSE ST 2.9 -7 -4 8 13 9 1 -2 5
TEXAS 2.9 -12 9 2 1 9 7 5 2
TEXAS A&M 2.9 -5 -2 10 7 9 6 9 -11
CLEMSON 2.8 -3 6 -1 13 2 9 -8 4
MINNESOTA 2.6 2 -1 12 -15 18 -1 7 -1
BOWLING GREEN 2.5 -2 4 0 0 1 8 15 -6
IOWA ST 2.5 6 4 4 -7 -5 14 10 -6
GEORGIA 2.4 10 -16 -3 9 -1 11 -2 11
TENNESSEE 2.4 4 3 2 7 1 -7 6 3
WISCONSIN 2.4 14 3 -8 -2 0 13 2 -3
NEVADA 2.3 5 -3 1 -3 12 0 1 5
RUTGERS 2.1 7 20 1 -6 11 -3 -7 -6
OKLAHOMA ST 2.0 12 0 5 -3 5 -15 16 -4
MIDDLE TENN 1.9 -19 12 3 11 6 6 -6 2
ARKANSAS 1.8 1 15 -9 4 -4 -1 -3 11
BALL ST 1.8 -5 -7 5 17 1 3 5 -5
E CAROLINA 1.8 -7 11 5 17 4 0 -4 -12
FLORIDA ST 1.8 4 1 0 6 -8 -4 7 8
STANFORD 1.8 13 0 -4 3 -11 4 8 1
TOLEDO 1.6 11 -10 0 1 -3 5 -2 11
LA TECH 1.5 -5 7 9 1 -8 13 0 -5
LOUISVILLE 1.3 3 -3 -12 0 2 6 11 3
TULSA 1.3 17 -2 -5 -6 -5 19 -11 3
W KENTUCKY 1.3 3 -12 -1 2 -4 4 8 10
C MICHIGAN 1.0 -11 5 4 6 7 5 -4 -4
ARIZONA ST 0.9 -6 -2 5 3 -1 7 5 -4
PENN ST 0.8 -4 6 7 2 1 3 -3 -6
AUBURN 0.6 5 1 -8 0 5 -3 4 1
KENTUCKY 0.6 -4 2 5 -1 15 -9 -2 -1
VIRGINIA 0.5 -7 0 -7 2 2 3 6 5
UAB 0.4 -6 9 -1 -5 6 5 0 -5
GEORGIA TECH 0.3 -6 8 2 -4 3 10 -13 2
MICHIGAN ST 0.3 5 -6 2 2 -1 -1 -10 11
FLORIDA ATLANTIC 0.3 -9 1 -9 18 1 -7 0 7
AKRON 0.1 -6 -5 6 -1 1 -7 6 7
NEW MEXICO 0.0 -12 -6 3 0 2 -1 9 5
OREGON ST -0.1 4 5 1 0 8 -14 3 -8
PURDUE -0.1 -6 -5 -3 1 -1 3 -2 12
SYRACUSE -0.1 -4 -6 -1 -6 11 -4 2 7
TROY -0.1 -1 -6 6 -1 -5 -9 11 4
INDIANA -0.3 -7 7 1 2 -4 -8 4 3
MICHIGAN -0.4 -10 -12 -10 2 14 5 6 2
ARIZONA -0.5 -4 -1 6 -1 7 -9 3 -5
KANSAS -0.5 -8 -5 4 21 -5 -8 4 -7
OHIO UNIVERSITY -0.5 -7 13 -12 2 -4 4 1 -1
VANDERBILT -0.5 -4 3 9 -2 -2 -2 4 -10
MIAMI OH -0.6 11 -24 -12 -3 0 13 -10 20
NORTHWESTERN -0.6 -1 4 -3 -9 -7 9 4 -2
UCF -0.8 4 8 1 1 -7 12 -3 -22
MEMPHIS -1.0 -13 -9 1 10 0 7 0 -4
KANSAS ST -1.1 4 7 -7 4 -4 -6 -6 -1
ARKANSAS ST -1.3 1 -5 9 -4 -4 5 -9 -3
HOUSTON -1.3 -6 4 -6 -7 9 -4 6 -6
MARYLAND -1.3 15 -6 -8 8 -6 -5 -9 1
TEXAS TECH -1.3 -3 -6 8 -5 0 9 -5 -8
UTEP -1.3 -5 -3 12 9 0 -10 2 -15
BYU -1.4 1 0 2 -6 14 2 -10 -14
COLORADO -1.5 0 -6 -7 -4 8 2 1 -6
MIAMI (FLA) -1.5 -8 0 -10 -6 -3 5 14 -4
USF -1.5 -3 0 -8 9 -4 5 -10 -1
TEMPLE -1.5 -6 4 0 0 -9 -8 2 5
BUFFALO -1.6 -8 -7 19 3 -4 -15 1 -2
S CAROLINA -1.8 0 -4 -11 -7 -2 2 1 7
W MICHIGAN -1.8 3 -6 -7 -5 12 6 -13 -4
CINCINNATI -2.0 -15 9 -8 16 -6 -7 -5 0
DUKE -2.0 -10 2 5 -3 -8 -8 2 4
E MICHIGAN -2.0 -11 -3 1 8 -6 10 -9 -6
NEBRASKA -2.0 -1 5 -11 -17 0 -2 -13 23
UCLA -2.0 -11 6 -10 -4 4 6 -6 -1
UTAH ST -2.0 -5 6 -1 2 -6 -2 -6 -4
LOUISIANA LFT (SW) -2.1 -3 3 -5 -5 -5 -4 -3 5
UL MONROE (NE) -2.1 -6 -5 2 1 10 1 -1 -19
KENT ST -2.3 0 -2 -3 -11 3 -11 -1 7
RICE -2.4 -7 -9 15 -2 8 -17 -4 -3
SAN DIEGO ST -2.9 -6 -10 -6 6 -7 2 -1 -1
HAWAII -3.0 12 -11 -7 -6 1 -3 -2 -8
MISS ST -3.4 7 -5 -4 0 -5 2 -1 -21
UNLV -3.4 -1 -3 0 1 -11 -8 -17 12
WASH ST -3.4 -1 -7 -25 -2 6 -5 -1 8
N CAROLINA -3.5 1 2 6 -6 -11 -1 -4 -15
WYOMING -3.5 -2 8 -22 -12 -4 -12 6 10
BAYLOR -3.6 0 -5 16 -18 -7 5 -15 -5
MISSISSIPPI -3.8 -6 -7 -2 -10 -3 -5 -3 6
COLORADO ST -4.4 -8 0 2 0 -7 5 -12 -15
N TEXAS ST -4.6 -3 -14 -17 -11 -15 0 13 10
N CAROLINA ST -4.8 7 -11 8 -16 -11 0 -17 2
FLORIDA INTERNATIONAL -4.8 4 -4 4 -14 -9 -8 -6 -5
FRESNO ST -4.9 -11 -10 -11 -9 -6 -4 7 5
ARMY -5.0 16 5 -14 -10 -18 -2 3 -20
IDAHO -5.4 2 -8 -14 -9 -1 -6 -2 -5
WASHINGTON -5.8 2 4 -17 -1 -8 -3 -19 -4
TULANE -6.3 -2 -16 -5 -3 -11 -12 -6 5
ILLINOIS -6.8 8 -4 -6 -2 -15 -11 -6 -18
MARSHALL -7.0 -6 -4 -4 -10 -9 -6 2 -19
SMU -7.3 -12 2 -13 -9 1 5 -19 -13
N MEXICO ST -9.1 -6 -10 -6 -15 -10 -23 5 -8

2010 Compu-Picks Blog

Questions, comments or suggestions? Email me at cfn_ms@hotmail.com

Follow cfn_ms on Twitter










Advertisement






Advertisement










Unauthorized use of ad tag