Turnover Analysis - Part 2: Do Turnovers = Turnaround?

Submitted by Enjoy Life on November 30th, 2009 at 5:46 PM
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-fareast-font-family:"Times New Roman"; mso-fareast-theme-font:minor-fareast; mso-hansi-font-family:Calibri; mso-hansi-theme-font:minor-latin;} table.MsoTableGrid {mso-style-name:"Table Grid"; mso-tstyle-rowband-size:0; mso-tstyle-colband-size:0; mso-style-priority:59; mso-style-unhide:no; border:solid black 1.0pt; mso-border-themecolor:text1; mso-border-alt:solid black .5pt; mso-border-themecolor:text1; mso-padding-alt:0in 5.4pt 0in 5.4pt; mso-border-insideh:.5pt solid black; mso-border-insideh-themecolor:text1; mso-border-insidev:.5pt solid black; mso-border-insidev-themecolor:text1; mso-para-margin:0in; mso-para-margin-bottom:.0001pt; 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;}

In Part 1(mgoblog.com/diaries/turnover-analysis-part-1-it-all-just-luck-1), a statistical analysis concluded: Luck is primarily responsible for TOM of approximately 80% of FBS football teams. For the other 20%, team performance (good or bad) is primarily responsible for TOM.

In Part 2, I’ll look at the question, “Do Turnovers = Turnaround”. Each year, Phil Steele includes his “Turnovers = Turnaround” article in his College Football Preview. A couple of quotes:

“Teams with a positive double-digit TO ratio had the same or weaker records 77% of the time since 1996.

“Teams with a negative double-digit TO ratio had the same or stronger records 80% of the time since 1996.”

Steele bases his analysis on the premise that turnovers are primarily luck and that teams “rarely get a repeat of that good fortune.”

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

Phil Steele is wrong. Turnovers do NOT equal turnaround.

The teams Steele isolates (those with double-digit turnovers) are the teams whose TOM is primarily due to performance and not luck. Therefore his basic premise is incorrect.

Even if the double-digit TOM was primarily due to luck, there is no cause and effect. The percentage of teams that “turnaround” the next season is approximately the same when TOM is completely ignored.

From 1999 to the present, 72% of all FBS teams that had a winning record of at least +2 (7-5 or better) had the same or weaker records the next year regardless of TOM. This includes approximately 50 teams each year. (Steele: 77% for the teams with double-digit turnovers.)

From 1999 to the present, 74% of all FBS teams that had a losing record of at least -2 (5-7 or worse) had the same or stronger records the next year regardless of TOM. This includes approximately 43 teams each year. (Steele: 80% for the teams with double-digit turnovers.)

Basically, it is very difficult for winning teams to keep on winning at the same rate and very difficult for losing teams to keep on losing at the same rate regardless of what TOM happens to be.

Problems With Analysis of TOM

It is very difficult to analyze TOM. It is undeniable that a single turnover can determine which team wins the game. It is also undeniable that teams win games with large negative TOM (Michigan had a -4 TOM and beat a favored Florida team in the 2007 Capital One Bowl). And, teams lose games with large positive TOM (Michigan lost to osu in 2006 with a +3 TOM).

Obviously, the impact of TOs is situational and difficult to analyze. But, using TOTAL TOM for a year in any analysis is foolhardy. Every football game is decided individually and TOs occur uniquely for each game. Averaging this data over a year just does not make sense.

This table shows the Top 10 teams from 2008, their Average TOM Per Year, and their yearly Win/Loss Record and yearly TOM for the last 8 years.

Team/Year

2002

2003

2004

2005

2006

2007

2008

Florida (+7.4)

8-5 (-9)

8-5 (+7)

7-5 (+4)

9-3 (+18)

13-1 (+5)

9-4 (+5)

13-1 (+22)

Utah (+7.7)

5-6 (-1)

10-2 (+9)

12-0 (+15)

7-5 (-1)

8-5 (+8)

9-4 (+11)

13-0 (+13)

USC (+13)

11-2 (+18)

12-1 (+20)

13-0 (+19)

12-1 (+21)

11-2 (+4)

11-2 (+2)

12-1 (+7)

Texas (+6.1)

11-2 (+17)

10-3 (+2)

11-1 (+5)

13-0 (+7)

10-3 (+9)

10-3 (+1)

12-1 (+2)

Oklahoma (+9.9)

12-2 (+19)

12-2 (+17)

12-1 (+4)

8-4 (-1)

11-3 (-1)

11-3 (+8)

12-2 (+23)

Alabama (+6.7)

10-3 (+15)

4-9 (+1)

6-6 (+6)

10-2 (+8)

6-7 (+7)

7-6 (+4)

12-2 (+6)

TCU (+8.1)

10-2 (+15)

11-2 (+4)

5-6 (+4)

11-1 (+21)

11-2 (+7)

8-5 (-7)

11-2 (+13)

Penn State (+2.6)

9-4 (+14)

3-9 (-6)

4-7 (-3)

11-1 (+3)

9-4 (+1)

9-4 (+2)

11-2 (+7)

Ohio State (+3.7)

14-0 (+13)

11-2 (+1)

8-4 (-1)

10-2 (-9)

12-1 (+9)

11-2 (-3)

10-3 (+16)

Oregon (+2.1)

7-6 (+5)

8-5 (-5)

5-6 (-2)

10-2 (+13)

7-6 (-10)

9-4 (+9)

10-3 (+5)

As you can see, there are wild swings in Average TOM Per Year for many of the teams with no apparent correlation to wins and losses.

A Bit More About Turnovers, Luck, and Michigan

In Part 1, I looked at a statistical analysis to evaluate whether turnovers are based on luck or team performance. It is also worthwhile to examine TOs individually and make a determination of whether luck or team performance is primarily involved. The criteria I used are shown in the table below.

Is the player doing what he should be doing and has been coached to do?

Does the TO occur rarely?

If the answers to these questions is “Yes”, the TO is primarily bad luck. If the answers are “No”, the TO is primarily due to poor performance. This year for M.

Player

Fumbles

Lost

Tate

11

5

Denard

5

3

Team (Those bad snaps)

3

0

Brown

3

1

Minor

2

1

Stonum

1

1

Mathews

1

1

Hemmingway

1

1

Odoms

1

0

Shaw

1

0

Total

29

13

This year Tate/Denard/Team fumbled the ball 19 times and lost 8. Almost all of these were due to poor performance and not bad luck (the lost fumble in the Iowa game, I would classify as bad luck because it was caused by weather). The good news is that these should be correctable.

In 2008, Jimmy Clausen had 17 interceptions. This year he has 4. Does anyone actually believe that is just due to bad luck last year and good luck this year? Most interceptions are performance based.

This year the freshman QBs threw 14 interceptions. These are mostly due to poor performance and should be correctable.

Last year, Michigan was -10 in TOM and ended up 3-9. This year, Michigan was -12 in TOM and ended up 5-7. Obviously a better TOM was not the cause of the “turnaround”. Why would TOs actually increase in the second year of the Coach Rod offense? Is this a major cause of concern?

The answer I think is obvious – freshmen QBs were the cause of the poor TOM!

 I will readily admit I was one of the MGoBloggers that did not listen when several of our fellow MGoBloggers warned us all that things would not end well with freshmen QBs. They were absolutely right, I was absolutely wrong. Many of the advances in team performance this year were masked by the inexperience and inconsistency of the freshmen QBs. This should not be a cause for concern since neither will be a freshman ever again.

Comments

michelin

November 30th, 2009 at 8:09 PM ^

As usual, I enjoyed reading your analysis. You are right that the analysis of turnovers is complex. In fact, my guess is that the FACT of having actual turnovers sometimes is less important than the FEAR of having turnovers. Although I don't recall how many turnovers Henne had during his freshman year, I do remember how Lloyd restricted the plays he was allowed to call--in large part because riskier throws would have resulted in a lot more turnovers. Those turnovers, certainly, could dramatically affect the outcome of a game. But they would never be recorded, since they were prevented by altering the offense. Those changes in the offense, of course, themselves reduced the chances of winning. So, it may appear that a poor offense was the cause of a poor record, when the fear of turnovers was partly responsible for the poor offense. So the FACT of having turnovers may be less important than the FEAR.

kman76

December 1st, 2009 at 12:42 AM ^

I especially liked the Time-Cube style graphics.

"Do you realize that a 4 corner square rotating 1/4 turn creates a full circle? A full rotated square will create 16 corners, 96 hours and 4 simultaneous 24 hour Day circles within only a single imaginary cubed Earth rotation."

stubob

December 1st, 2009 at 12:54 AM ^

I think the graph of the annual team's record and the difference from an average season (6-6) is more interesting than the turnover analysis. I was looking at turnover stats earlier this month as well, and I'll add my two cents.

1. Very few teams fall outside the statistical noise range. Out of 120+ teams, half have a turnover margin between .5 and -.5 per game. Only 4 have a margin greater than 1.5 or less than -1.5. That kind of distribution makes it hard to make much of a definitive claim.

Chart of the distribution of turnover margin for D-1:

Doesn't get much more of a bell-curve than that.

2. I also expected interceptions to be less random than fumbles. I don't think the data shows it. You may be able to claim that the fumble data is a little noisier and the interception data is more linear, but not by much. It's still in the range of +- 5, over 10 games.

Chart of smoothing give-away/take-away for fumbles and interceptions (2 and 3-number average to try and show some trending):

3. Looking for patterns to be able to say "Good teams (highly ranked teams) have fewer turnovers (higher ranked turnover margins) than bad teams" isn't shown by the data either. My conclusive evidence is that Georgia, #118 in turnover rank, is ranked #33. Rutgers, #1 in turnover rank, is ranked #35. I think it's safe to say "Good teams don't turn over the ball as much as bad teams," but the converse "A low turnover margin makes a good team" isn't true.

Chart of the top 40 D-1 teams, and the delta between rank and turnover rank:

So, I think that you're right, in saying that TO margin isn't a major factor in the overall record of a team. But I think the explanation is simpler than what you've put out. Even a +- 10 TO margin is only +-1 per game, which is likely to be insignificant.

However, in 2007 Clausen had 6 picks on roughly half the throws. So his freshman year would be comparable, if not better than his sophomore year. We're looking at 6 out of 240, or 17 out of 450, or 4 out of 420. It's noise. Look at McCoy's numbers. I'd say about the same year-to-year for interceptions. Same with Henne. It's just a percentage of throws, with some statistical variance thrown in.

Enjoy Life

December 1st, 2009 at 12:30 PM ^

I think we are basically saying the same thing. My analysis shows that 80% of teams fall within the noise range. That means only about 12 good teams and 12 bad teams fall outside the "luck" quotient.

However, saying that a TOM of +/- 10 is less than 1 per game is mathematically correct but when I have looked at game by game TOM, there are usually many games with insignificant TOM but also one or two with TOM of +/- 3-5. So, TOM does not affect probably 80% of games but the 20% that are affected can have a dramatic impact.

For example, a 6-6 team turns into an 8-4 or 4-8 team if those 2 games are + or - in large TOM.

oakapple

December 1st, 2009 at 8:58 AM ^

Michigan's terrible turnover margin was influenced by two factors.

The first was the use of Denard Robinson. No one denies that Robinson has great talent, but he was a turnover machine. In relation to the amount of time he played, Robinson produced turnovers at a prodigious pace. Forcier's turnover rate, in relation to the amount of time he played, was a lot more reasonable (though still not acceptable).

The other factor was losing David Molk. I suspect that the bad-snap turnovers were mainly attributable to that.

michelin

December 1st, 2009 at 9:49 AM ^

To test the hypothesis that bad offensive teams restrict their playbook to avoid turnovers but still have a high number of turnovers per play, try dividing the number of fumbles for the number of plays (or a surrogate measure like yards gained, which is probably easier to obtain).

Another interesting possiblity is to do a sum/difference plot, where you look at the number of turnovers divided by the number of plays on the x axis to see whether the offense is error-prone. On the y-axis plot the product X*Y, to look at whether the offense if high-risk high-reward and has a lot of plays but also a lot of turnovers. You may see some trends or unexpected results.

Procumbo

December 3rd, 2009 at 2:28 PM ^

We agree that Denard's interceptions aren't random right? I mean, when you lob it to a defender with no offensive players in the area, that results in an interception a lot of the time. Good quarterbacks don't do that. Denard does it a lot.