Normal 0 false false false EN-US X-NONE X-NONE /* Style Definitions */ table.MsoNormalTable {mso-style-name:"Table Normal"; mso-tstyle-rowband-size:0; mso-tstyle-colband-size:0; mso-style-noshow:yes; mso-style-priority:99; mso-style-qformat:yes; mso-style-parent:""; mso-padding-alt:0in 5.4pt 0in 5.4pt; mso-para-margin-top:6.0pt; mso-para-margin-right:0in; mso-para-margin-bottom:6.0pt; mso-para-margin-left:0in; mso-pagination:widow-orphan; font-size:11.0pt; font-family:"Calibri","sans-serif"; mso-ascii-font-family:Calibri; mso-ascii-theme-font:minor-latin; mso-hansi-font-family:Calibri; mso-hansi-theme-font:minor-latin;}

A comprehensive analysis of turnovers is a bit too much to cover all at once. So, I’ve decided to break it up into a few parts. Part 1: Are Turnovers Just A Matter of Luck?

Some folks claim that turnovers are primarily a matter of luck and that teams have little or no control over turnover margin (TOM). Phil Steele is one of the most notable advocates that turnovers are primarily luck. Each year, Steele includes his “Turnovers = Turnaround” article in his College Football Preview. A couple of quotes:

“Teams that benefitted from double-digit turnovers the previous year rarely get a repeat of that good fortune.”

“Let’s take a look at some teams who had terrible luck (lots of turnovers) in one year and then drastically improved over the next year without those turnovers.”

In Part 2 of the Turnover Analysis, I’ll look at Steele’s theory about turnovers being a cause of turnarounds. But, for now, let’s just look at whether turnovers are primarily luck.

Let’s first define the term: “Luck is a belief in good or bad fortune in life caused by chance which happens beyond a person's control.” As applied to turnovers, this would mean they simply happen at random (i.e. chance) and a football team has no control over TOM.

Executive Summary: The gory details are below but for those in need of instant gratification here is the synopsis:

Disclaimer: There is obviously an element of luck and an element of skill involved in the sport of football. As you’ll see, the analysis is to determine the “primary” cause of turnovers. It is not attempting to conclude that turnovers are completely luck or completely skill.

Basis: All 120 FBS teams over the last 10 years (1999 through 2008); Total TOM Per Year over the last 10 years. Bowl games excluded before 2002.

Here is a table of the very good teams and very poor teams with their Average TOM per year over the last 10 years and their Average WLM (Win/Loss Margin) over the same 10 years. Similar to TOM, the win/loss margin is merely games won minus games lost. For example, a team that is 7-5 has a WLM of +2 and a team that is 5-7 has a WLM of -2. I decided to use WLM because it provides data that is in the same format as TOM (i.e. net numbers).

Table Showing Very Good and Very Poor Teams in the FBS
Average TOM/Yr and Average WLM/Yr: 10 Years (1999 through 2008)
		TOM	WLM				TOM	WLM
Team	CONF	AVG	AVG		Team	CONF	AVG	AVG
USC	PAC10	10.2	7.1		Florida Intl	SunBelt	-5.1	-5.4
Oklahoma	Big12	8.1	8.4		Utah St	WAC	-5.4	-5.0
West Virginia	BigEast	7.8	3.8		Baylor	Big12	-5.8	-4.7
Virginia Tech	ACC	7.7	7.1		Idaho	WAC	-7.9	-5.6
TCU	MW	7.4	6.0		SMU	CUSA	-8.4	-5.5
Texas	Big12	7.3	8.4		Army	Army	-10.1	-7.0
Florida	SEC	6.1	6.4
Utah	MW	5.2	4.6
Boise St	WAC	4.9	8.5
Boston College	ACC	4.7	4.8
Georgia	SEC	4.7	6.7
Michigan	Big10	4.5	4.7
Florida State	ACC	4.2	5.1
Oregon	PAC10	4.1	4.6

And, yes, that does say that USC has averaged over +10 TOM Per Year for the last 10 years (that includes one year at -19 TOM, one year at +21, and six years with double digit positive TOM).

The Gory Details – TOM Simulation

To determine if TOM was primarily due to luck, I designed a simulation to provide TOM data that was based entirely on luck. The simulation is based on rolling 2 dice. Rolling dice involves random, independent events and the results are based purely on luck. Instead of adding the numbers on the two dice (craps), the numbers are subtracted.

One die is red for TO Lost and one is green for TO Gained. Obviously, TOM = Green – Red. Thus, the maximum TOM would be +/- 5 per game which is very consistent with actual data. Also, the distribution curve of turnovers is bell shaped with the most likely value being 0 (17% Chance) and least likely being 5 (3% Chance for + and 3% chance for -). See the table below for a comparison of the simulation distribution curve versus the theoretical curve.

Actually rolling the dice enough time to get statistically meaningful data would have taken way too long and would be prone to error. So, I used an EXCEL spreadsheet to accomplish the same results. EXCEL has formulas to generate random numbers within a range. This actually works better than the dice because I could set the lower range at -0- (the fewest possible turnovers a team could experience in a game) and the higher range at +5 (the most possible turnovers a team would experience in a game).

 I created a formula to subtract one random number from a second random number which results in TOM per game. I then created a table with 12 columns (one for each game in a year) and 1200 rows (120 FBS teams over 10 years). Each time the F9 key is pressed, the random numbers and TOM are recalculated for all 120 teams over 10 years (14,400 games). I used 10 trials and took the composite of all the trials (144,000 games). The composite is based on a count of the number times each TOM occurs.

Here is a table showing all possible TOM for a game, the % that occurred in the simulation, and the theoretical % that should occur. This demonstrates the validity of the simulation.

TOM	SIM%	THRY%
5	2.7%	2.8%
4	5.6%	5.6%
3	8.3%	8.3%
2	11.0%	11.1%
1	13.8%	13.9%
0	16.7%	16.7%
(1)	14.0%	13.9%
(2)	11.1%	11.1%
(3)	8.5%	8.3%
(4)	5.5%	5.6%
(5)	2.8%	2.8%

I’ve worked with simulations involving dice before and expected some variation but, I would have to say, these results shocked me.

Simulation Results: Average for 10 Trials
120 Teams: 10 Years, 12 Games Per Year
TOM/YR	%	NO.
9+	0.1%	0
8 to 8.99	0.1%	0
7 to 7.99	0.1%	0
6 to 6.99	0.3%	0
5 to 5.99	1.5%	2
4 to 4.99	4.3%	5
3 to 3.99	7.2%	9
2 to 2.99	8.5%	10
1 to 1.99	13.7%	17
0 to .99	12.9%	16
0	1.4%	2
0 to -.99	15.2%	18
-1 to -1.99	12.5%	15
-2 to -2.99	10.0%	12
-3 to -3.99	5.7%	7
-4 to -4.99	3.5%	4
-5 to -5.99	2.3%	3
-6 to -6.99	0.3%	0
-7 to -7.99	0.1%	0
-8 to -8.99	0.2%	0
-9+	0.1%	0
	100.0%	120

Based on these simulation results, turnovers could be explained as primarily luck even with several teams experiencing up to +/- 9 Average TOM/Year over 10 years. Note that in the simulation, approximately 90% of all FBS team’s average between approximately +/- 4 turnovers over 10 years. The detailed simulation data also shows that, by just luck, many teams could experience double-digit turnovers in multiple years.

I also ran the simulation for a span of 100 years for each team. As expected the variation was reduced significantly. Approximately 80% of all teams had an average TOM/YR of less than +/- 1.0 and 100% had an average TOM/YR of less than +/- 2.0.

Here are several examples of actual data from the simulation (all examples are from the 10 year simulation).

Example 1: Actual Data from the Simulation (Large Negative Average TOM/YR)
Game-->	1	2	3	4	5	6	7	8	9	10	11	12	TOM	AVG
Year 1	(2)	3	(4)	(1)	(1)	(4)	(1)	(3)	(4)	(4)	(2)	3	(20.0)
2	(2)	3	(1)	(4)	0	1	(2)	(1)	(3)	(2)	0	(2)	(13.0)
3	4	(2)	(2)	(3)	2	1	0	(2)	(2)	0	(4)	(1)	(9.0)
4	2	0	0	(1)	(4)	3	(2)	(3)	2	3	2	3	5.0
5	(3)	(2)	(4)	0	(1)	1	0	3	5	(3)	4	(1)	(1.0)
6	(1)	2	(2)	(1)	4	(1)	0	(2)	(4)	4	3	(2)	0.0
7	2	(3)	0	(3)	(2)	3	(1)	0	(3)	(4)	0	0	(11.0)
8	0	2	2	(1)	(4)	(3)	5	(4)	(5)	(4)	(2)	(3)	(17.0)
9	1	(3)	5	1	(3)	3	(1)	(1)	(2)	(3)	(4)	(2)	(9.0)
10	(1)	0	0	1	1	3	(5)	0	(4)	1	(3)	0	(7.0)	(8.2)

 

Example 2: Actual Data from the Simulation (Large Positive Average TOM/YR)
Game-->	1	2	3	4	5	6	7	8	9	10	11	12	TOM	AVG
Year 1	1	5	4	(2)	(2)	2	0	0	(2)	0	0	1	7.0
2	(1)	2	1	0	(1)	3	(4)	4	4	(2)	(2)	0	4.0
3	4	2	3	2	3	1	(3)	1	(1)	(1)	(5)	1	7.0
4	2	3	4	(2)	(4)	3	(1)	(4)	1	(4)	0	1	(1.0)
5	5	1	3	3	2	4	0	(1)	(1)	(1)	(3)	2	14.0
6	0	5	3	3	(1)	(3)	(3)	0	(1)	3	2	0	8.0
7	2	(1)	0	0	(1)	3	2	2	4	(1)	0	4	14.0
8	0	5	3	1	(2)	(1)	(2)	0	(2)	(2)	2	1	3.0
9	0	3	2	(2)	2	1	0	0	5	(1)	1	3	14.0
10	3	5	1	4	1	2	0	3	2	0	1	1	23.0	9.3

 

Example 3: Actual Data from the Simulation (Average TOM/YR Approximately -0-)
Game-->	1	2	3	4	5	6	7	8	9	10	11	12	TOM	AVG
Year 1	1	2	0	4	(5)	(2)	3	0	0	(5)	(4)	0	(6.0)
2	(1)	0	1	1	0	4	(2)	(1)	2	0	(2)	3	5.0
3	2	(4)	0	4	(3)	0	5	(1)	1	(3)	(1)	1	1.0
4	5	(1)	5	(3)	0	0	(3)	4	(2)	(5)	(3)	0	(3.0)
5	(2)	4	0	0	(2)	3	0	0	(2)	3	1	1	6.0
6	(3)	(2)	(3)	0	(1)	4	(3)	4	(3)	0	0	2	(5.0)
7	1	(3)	(1)	0	(2)	1	3	(2)	4	(1)	(1)	0	(1.0)
8	(4)	(3)	(1)	1	1	3	2	(5)	3	5	0	(3)	(1.0)
9	3	3	(1)	(4)	1	1	3	(1)	1	2	4	2	14.0
10	(1)	(2)	1	(3)	0	0	(2)	0	5	0	2	(3)	(3.0)	0.7

 

The Gory Details – Actual Data

So, what does the actual data show? I looked at all FBS teams from 1999 to 2008. I tracked turnover margin (TOM) and win/loss margin (WLM).

Even though the simulation indicates a relatively large variation would be expected in TOM even if only luck is involved, a significant number of teams fall outside of the expected variation. Here is a table showing all the teams with average TOM per year greater than 4.0 (sorted by TOM).

Table Showing All Teams With Average TOM/Year Greater Than 4.0 (Sorted by TOM)
Team	CONF		Avg	1999	2000	2001	2002	2003	2004	2005	2006	2007	2008
USC	PAC10	TOM	10.2	14	(19)	16	18	20	19	21	4	2	7
		WLM	7.1	0	(2)	0	9	11	13	11	9	9	11
Oklahoma	Big12	TOM	8.1	(4)	6	10	19	17	4	(1)	(1)	8	23
		WLM	8.4	3	12	8	10	10	11	4	8	8	10
West Virginia	BigEast	TOM	7.8	(5)	7	(8)	19	16	3	14	7	13	12
		WLM	3.8	(3)	1	(5)	5	3	4	10	9	9	5
Virginia Tech	ACC	TOM	7.7	3	6	10	8	(1)	13	9	4	11	14
		WLM	7.1	11	9	5	6	3	7	9	7	8	6
TCU	MW	TOM	7.4	4	10	3	15	4	4	21	7	(7)	13
		WLM	6.0	3	9	1	8	9	(1)	10	9	3	9
Texas	Big12	TOM	7.3	11	8	11	17	2	5	7	9	1	2
		WLM	8.4	5	7	8	9	7	10	13	7	7	11
Wake Forest	ACC	TOM	6.3	6	(9)	(3)	18	7	7	(2)	13	9	17
		WLM	0.4	1	(7)	1	1	(2)	(3)	(3)	8	5	3
Florida	SEC	TOM	6.1	(6)	19	(4)	(9)	7	4	18	5	5	22
		WLM	6.4	6	8	7	3	3	2	6	12	5	12
S. Mississippi	CUSA	TOM	5.6	10	0	7	(3)	5	5	14	6	(1)	13
		WLM	2.5	5	3	1	1	5	2	2	4	1	1
W. Kentucky	SunBelt	TOM	5.3				17	10	8	3	(4)	2	1
		WLM	2.3				9	5	6	1	1	2	(8)
Toledo	MAC	TOM	5.2	8	22	3	7	11	(2)	5	(3)	1	0
		WLM	2.6	1	9	7	4	4	5	6	(2)	(2)	(6)
Utah	MW	TOM	5.2	8	(11)	1	(1)	9	15	(1)	8	11	13
		WLM	4.6	5	(3)	3	(1)	8	12	2	3	5	12
Air Force	MW	TOM	5.1	(4)	7	8	9	6	1	(7)	8	10	13
		WLM	1.1	1	5	0	3	2	(1)	(3)	(4)	5	3
Boise St	WAC	TOM	4.9	10	8	(8)	8	10	10	(8)	11	1	7
		WLM	8.5	5	7	4	11	12	10	5	13	7	11
Boston College	ACC	TOM	4.7	2	11	3	8	3	0	(4)	15	6	3
		WLM	4.8	5	1	3	5	3	6	6	7	8	4
Georgia	SEC	TOM	4.7	8	(1)	1	8	11	(2)	11	(1)	9	3
		WLM	6.7	3	3	5	12	8	8	7	5	9	7
Alabama	SEC	TOM	4.7	4	(8)	4	15	1	6	8	7	4	6
		WLM	2.3	8	(5)	1	7	(5)	0	7	(1)	1	10
Michigan	Big10	TOM	4.5	10	11	(4)	9	2	6	5	14	2	(10)
		WLM	4.7	7	5	5	7	7	6	2	9	5	(6)
Texas A&M	Big12	TOM	4.5	4	6	3	2	(11)	9	6	9	7	10
		WLM	1.0	5	3	3	0	(4)	2	(1)	5	1	(4)
Florida State	ACC	TOM	4.2	8	10	4	11	8	7	(4)	(8)	6	0
		WLM	5.1	11	10	3	4	7	6	3	1	1	5
Oregon	PAC10	TOM	4.1	9	3	14	5	(5)	(2)	13	(10)	9	5
		WLM	4.6	6	7	9	1	3	(1)	8	1	5	7

 This table includes 21 teams. However, as the simulation indicates, approximately 7 teams should have TOM greater than 4 if luck is primarily responsible. So, 7 of these teams needed to be eliminated. I decided to use low WLM as the criteria to eliminate teams. Teams that are eliminated and their WLM are: Wake Forest (0.4), Texas A&M (1.00), Air Force (1.1), Alabama (2.3), Western Kentucky (2.3), Toledo (2.6), and S. Mississippi (2.5). That leaves the 14 teams in the summary table included above in the Executive Summary.

Here is a table showing all the teams with average TOM per year less than negative 4.0 (sorted by TOM).

Table Showing All Teams With Average TOM/Year Less Than Negative 4.0 (Sorted by TOM)
Team	CONF		Avg	1999	2000	2001	2002	2003	2004	2005	2006	2007	2008
Kent State	MAC	TOM	(4.2)	(11)	(2)	3	(16)	7	(1)	(11)	3	(11)	(3)
		WLM	(4.5)	(7)	(9)	(1)	(6)	(2)	(1)	(9)	0	(6)	(4)
Wyoming	MW	TOM	(4.6)	2	(9)	(3)	(2)	10	6	(12)	(4)	(12)	(22)
		WLM	(3.2)	3	(9)	(7)	(8)	(4)	2	(3)	0	(2)	(4)
Illinois	Big10	TOM	(5.0)	13	(2)	5	(8)	(18)	(6)	(11)	(15)	(2)	(6)
		WLM	(1.8)	3	(1)	9	(2)	(10)	(5)	(7)	(8)	5	(2)
Florida Intl	SunBelt	TOM	(5.1)				2	(5)	(6)	(8)	(9)	(14)	4
		WLM	(5.4)				(1)	(8)	(4)	(1)	(12)	(10)	(2)
Utah St	WAC	TOM	(5.4)	(11)	(2)	(13)	(11)	(4)	(6)	(2)	(6)	2	(1)
		WLM	(5.0)	(3)	(1)	(3)	(3)	(6)	(5)	(5)	(10)	(8)	(6)
Rutgers	BigEast	TOM	(5.7)	(5)	(7)	(22)	(13)	(6)	(7)	(3)	11	(6)	1
		WLM	(1.9)	(9)	(5)	(7)	(10)	(2)	(3)	2	9	3	3
Baylor	Big12	TOM	(5.8)	(5)	(9)	(3)	(17)	(5)	(15)	5	(7)	(18)	16
		WLM	(4.7)	(9)	(1)	(5)	(6)	(6)	(5)	(1)	(4)	(6)	(4)
Washington St	PAC10	TOM	(5.8)	(1)	(3)	(3)	1	(4)	(19)	(3)	(8)	(1)	(17)
		WLM	(1.8)	3	9	5	1	0	(10)	(7)	(2)	(5)	(12)
New Mexico St	WAC	TOM	(6.1)	6	(5)	(5)	0	(8)	5	(23)	(10)	(15)	(6)
		WLM	(3.8)	1	(5)	(2)	2	(6)	(1)	(12)	(4)	(5)	(6)
N. Carolina	ACC	TOM	(6.7)	2	(12)	(11)	(15)	(15)	(4)	(1)	(11)	(6)	6
		WLM	(2.4)	(5)	1	2	(6)	(8)	0	(1)	(6)	(4)	3
Idaho	WAC	TOM	(7.9)	0	(12)	(16)	(14)	(5)	(2)	(6)	(1)	(9)	(14)
		WLM	(5.6)	3	(1)	(9)	(8)	(6)	(6)	(7)	(4)	(10)	(8)
SMU	CUSA	TOM	(8.4)	(4)	(13)	(7)	(12)	(13)	(19)	5	1	(9)	(13)
		WLM	(5.5)	(2)	(6)	(3)	(6)	(12)	(5)	(1)	0	(10)	(10)
Army		TOM	(10.1)	(4)	(6)	(16)	(14)	(20)	3	(2)	(18)	(10)	(14)
		WLM	(7.0)	(5)	(9)	(5)	(10)	(13)	(7)	(3)	(6)	(6)	(6)

This table includes 13 teams. However, as the simulation indicates, approximately 7 teams should have TOM less than negative 4 if luck is primarily responsible. So, 7 of these teams needed to be eliminated. I used high WLM as the criteria to eliminate teams. Teams that are eliminated and their WLM are: Illinois (-1.8), Washington St (-1.8), Rutgers (-1.9), N. Carolina (-2.4), Wyoming (-3.2), New Mexico State (-3.8), and Kent State (-4.5). That leaves the 6 teams in the summary table included above in the Executive Summary.

In Part 2 of the Turnover Analysis, I’ll look at Steele’s theory about turnovers being a significant cause of turnarounds. I’ll also discuss why turnovers are (or aren’t?) important.