The Real Truth About Turnovers and Winning

Submitted by Enjoy Life on

Regardless of What We Have Been Told Or What We May Want To Believe, Here's The Bottom Line: For Over 80% Of All College Football Games, Turnover Margin Is NOT A Significant Factor In Determining Which Team Wins!

(BTW, Since turnover margin does impact which team wins in 20% of all football games, it is important ---- it's just not nearly as important as we have been lead to believe.)

Truthiness: "Truthiness is what you want the facts to be, as opposed to what the facts are. What feels like the right answer as opposed to what reality will support." (Stephen Colbert, October 17, 2005 – The Colbert Report). "The quality of preferring concepts or facts one wishes to be true, rather than concepts or facts known to be true" (American Dialect Society, January 2006).

No Way!: Uh yeah, we have all been wrong (myself included). I started looking at turnovers (TOs) in detail after the 2008 season when Michigan went 3-9 with a turnover margin (TOM) of –10. I wrote a series of diaries that concluded double digit TOs were caused primarily by the skill & experience of a team and not primarily by luck. Thus, good teams tend to have positive TOMs and poor teams tend to have negative TOMs. This is basically the opposite of believing that TOs are a primary factor in determining whether a team is good (i.e. winning record) or poor (i.e. losing record) – sorry, Phil Steele.

I also concluded that any analysis using the total TOs for an entire season was misleading, irrational, and just plain lazy (sorry again, Phil Steele). TOs must be analyzed on a game-by-game basis ONLY. After a game-by-game analysis is completed, it is valid to summarize the results for an entire season.

Then, a few weeks ago when I wrote the diary "Turnovers - The Year In Review", I started to see a trend in the turnover data that seemed pretty weird. Although M had a TOM of –9 for the year, the game-by-game situational review indicated that: "In reality, positive TOMs helped Michigan win as many games this year as negative TOMs contributed to M losing games!"

imageSeveral comments on that blog post did not agree with my conclusions (surprise!?) with psychomatt providing a table showing the relationship between "End-of-year TOM" for 40 teams and "Win/Loss Record" (this sample represents 33% of all FBS teams!). Based on this data, it sure looks like TOs could account for an average of a 5 win difference with an average TOM/Year difference of 23. A simple calculation (5 wins/23 TOM) yields the result that every TO gained is worth 0.217 games! (BTW, a similar approach has been used to determine the average value of a TO in the NFL is 0.207 games.)

Therefore, if Michigan had ended the year with a +13 TOM rather than a –9 TOM, the difference of 22 in TOM would result in 4.78 more wins and Michigan would have been expected to have a 12-0 (or at least an 11-1) season this year (woooo, hooooo!). But, is this really true or just truthiness?

Since this is a slow time of year, I decided to bite the bullet and take a more detailed look at the relationship between TOM and which team wins a football game.

In reality, the average value of a TO is 0.094 additional games won. Thus, the average difference between a team with one of the best TOMs and one of the worst TOMs is 2.4 additional games won during the season. (This is less than half of the impact that is usually attributed to TOs.) So, Michigan would have been expected to have a 9-3 or 10-2 season if the TOM had been +13 rather than –9.

The Methodology: One of the most difficult aspects of analyzing the impact of TOs is establishing how much each TO is worth. Luckily the folks at Football Outsiders have done an in-depth analysis to determine an average TO is worth approximately 4 points. [EDIT: This is a "swing" of 4 points. As explained in the comments below, the proper way to look at this is that the team committing the TO loses 2 points and the team getting the takeaway gains 2 points.]

So, that is what I used. Only if the winning team had a positive TOM and the margin of victory was within the value of TOM (4 X TOM), was the game counted in "turnover margin was a significant factor in determining which team wins". If the winning team had a positive TOM but the margin of victory was outside the value of TOM, then the game was NOT counted since TOs may have impacted the margin of victory (making it larger) but were not a significant factor in determining which team wins. For example, Michigan beat ND 28-24 with a TOM of +3 (value = 12 points). Since the margin of victory was within the value of TOM (12 points), this game was counted as TOs being a significant factor in determining the winning team. However, Michigan beat UConn 30-10 with a TOM of +1 (value = 4 points). Since the margin of victory was outside the value of TOM (4 points), this game was NOT counted as TOs being a significant factor in determining the winning team.

I looked at every game for the following three categories of teams: (1) Top 10% (12 teams) of End-of-Year TOM; (2) Bottom 10% (12 teams) of End-of-Year TOM; (3) Middle 10% (12 teams) of End-of-Year TOM. This is a total of 30% of all FBS teams and comprises 438 total games played.

imageThe Results: As the table indicates: (Column 2) 21% of all games end with a TOM of –0-; (Column 3) 17% of all games are won by the team with a negative TOM; (Column 4) 44% of all games are won by the team with a positive TOM but the TOs were not a significant factor in determining the winning team; and (Column 5) this totals to an average of 82% of all football games where TOM is not a significant factor in determining the winning team.

 

 

What About, "80% of the time the team that wins the TO battle, wins the football game"?: Some version of this statement is repeated hundreds of times during every football season. This statement is technically true. But, this statement is also worded very carefully and is a classic example of a contextual lie (stating part of the truth out of context, knowing that without complete information, it gives a false impression).  We expect such shenanigans in politics but it is quite annoying in sports.

imageThis table shows how the statement is true but let's dissect the statement to see how it is misleading. First, there is no mention of how many football games are excluded (i.e. all games that end with a TOM of –0-). As shown above, about 21% of all games end with neither team winning the TO battle. Second, there is no mention of the fact that many games are so non-competitive that TOs could not possibly have impacted which team wins. Thus, many folks who hear this statement are left with the false impression that TOs determine which team wins 80% of all football games! As shown above, this is overstated by a factor or 4 (TOs determine which team wins in only 18% of all football games).

Also note that if the basis is total games played (rather than just games that end up without TOM = 0), only 62% of games are won by the team that wins the TO battle.

 

 

 

 

Other Interesting Results: Some other results from the analysis.

Of the 12 teams with the largest positive TOM:

Average net wins per team was just 1.5

2 teams received an advantage of 5 net wins due to TOM: Oklahoma (11-2) and Toledo (8-4)

1 team received an advantage of 3 net wins: Maryland (8-4)

1 team received an advantage of 2 net wins: Army (6-6)

5 teams received an advantage of 1 net win: Virginia Tech (11-2), Hawaii (10-3), osu (11-1), Oregon (12-0), and Tulsa (9-3)

3 teams received no net wins: Stanford (11-1), Wisconsin (11-1), and Iowa (7-5)

4 teams actually lost a game due to negative TOM for that particular game: Virginia Tech, Army, Iowa, and Tulsa

 

Of the 12 teams with the largest negative TOM:

Average net losses per team was just 0.92

2 teams received a disadvantage of 3 net losses: Texas (5-7) and Duke (3-9)

2 teams received a disadvantage of 2 net losses: Eastern Mich (2-10) and Central Mich (3-9)

2 teams received a disadvantage of 1 net loss: Cincinnati (4-8) and Fresno State (8-5)

5 teams received no net losses: Middle Tenn (6-6), Memphis (1-11), New Mexico (1-11), UCLA (4-8), and Michigan (7-5)

1 team received an advantage of 1 net win with no net losses and 1 net win due to positive TOM for that particular game: Fla. Atlantic (4-8)

3 teams actually won 2 games due to positive TOM for those games: Middle Tenn, Duke, and Central Mich

4 teams actually won 1 game due to positive TOM for that game: Memphis, New Mexico, Michigan, and Fla. Atlantic

 

 

 

 

Sample Data: Finally, here are 4 detailed samples of the data. IMP = Impact (Yes, No, Opposite) Opp indicates the team that won the TO battle, lost the game!

image

image

Comments

jshclhn

December 30th, 2010 at 8:47 PM ^

I personally don't like the presumption that one factor (turnovers or otherwise) is responsible for X% of wins.  There are many instances where I would say no one factor was responsible for a win or a loss, but the sum of a couple factors.

Let's take this last Michigan season.  I don't think it's a coincidence that we had a positive turnover margin to start the season, then once we hit the meat of the Big Ten season, the TOM went south along with our winning percentage.  While we are at, we could look at turnover margin in 2008 and 2009 and how neither the TOM or winning percentage was exactly where we wanted it to be.  I would take that as very strong anecdotal evidence that turnover margin is pretty important.  However, TOM doesn't take into account a generally crappy offense (2008) or defense (2010 especially), so there are other factors that play into a win or loss.  

Granted, I didn't take mountains of actual data into account or put a nice percentage on it so take my analysis for what it's worth.

Also, I would like to know how a TOM of 5 or 7 in the tables immediately above can't be considered significant, especially when there's games when some teams don't even have 10 possessions.

Enjoy Life

December 30th, 2010 at 11:08 PM ^

I was isolating those turnovers that had a high probability of impacting which team wins the game versus those TOs that probably only made the margin of victory greater.

Wisconsin beat Northwestern 70-23 with a TOM of +7. The TOs were significant to the final score and margin of victory but each TO would have to be worth virtually 7 points to have been a determining factor in the fact that Wisconsin won the game. Even with a TOM of 0, Wisconsin would still have won.

Same with Texas. They lost to Kansas St by 14-39 with a TOM of -5. Each TO would need to be worth over 6 points to have an impact of Texas losing the game. In all likelihood, the TOs just meant Texas lost by a larger margin. Even with a TOM of 0, Texas would still have lost.

So, I was using the term significant only in the context of impacting which team wins.

bjk

December 31st, 2010 at 2:22 AM ^

but not before I had to go to Football Outsiders in an attempt to understand how they went about valuing a turnover at about 4 points. I came to a full stop here:

The ball on the opposing yard line is worth 6 points, leaving out the extra point which is pretty much guaranteed anyway.

My intuition is, if it is "pretty much guaranteed anyway," wouldn't you include it? They seem to imply that you get the extra point whether you score the 6 points or not.

I will try to read all this again when I am refreshed. Good post.

Enjoy Life

December 31st, 2010 at 1:38 PM ^

Sorry for the confusion (I have edited the OP to reflect the following).

The 4 points per TO is actually the "swing" in points. And, you are correct that the proper way to look at this is that the team commiting the giveaway ends the game with 2 points less than they would have (w/o any TOs). The team getting the takeaway ends the game with 2 points more than they would have (w/o any TOs).

It is based on expected points (EP) that the Mathlete http://mgoblog.com/diaries/never-punt-denard-fourth-down-strategy-revisited , Football Outsiders, and others use.

If you have the ball on your own 30 yard line the EP = 1.8 points. So, a turnover costs your team 1.8 points.

But, it also gives the ball to the other team on your 30 yard line which is worth 3.8 EP. However, the other team was always going to get another possession no matter what happens (even if you score a TD) so we have to subtract the value the other team would have received if the TO had not occurred. A good approximation is that the other team would have gotten the ball 40 yards further away (based on a net punt of 40 yards). Thus, the other team would have gotten the ball on their own 30 (rather than your 30) and would have had an EP of 1.8 from there. Thus the difference is 3.8-1.8 = 2

These values get rounded to +2 and -2 for a 4 point swing.

For example: Wisconsin 70-23 with +7 TOM. Without any TOs the score would have been Wisconsin 56 (70-14) and Northwestern 37 (23+14).

My name ... is Tim

December 31st, 2010 at 4:58 PM ^

This was an interesting analysis. However, I think one of the problems with a statistical average of turnovers is that it ignores the situational aspect that each one has (not that you haven't provided that caveat, as your prior post certainly took this to heart). I agree with your conclusion that the TOs likely didn't have that huge of an impact on our season, but part of that has to do with the fact that we likely won as many games as we deserved to win this season (meaning I would have expected us to win those 7 games with a TOM of 0, and lose those 5 games with a TOM of 0). I guess my concern going forward would be next year, when our offense and defense have (presumably) improved, are we still going to be unable to force turnovers to win us games, and are we going to turn the ball over and cost ourselves games against teams that we are now presumably equal to or better than.

trackcapt

December 31st, 2010 at 6:11 PM ^

So instead of 12-0 we would've been 9-3?  I'm guessing those 3 losses would've been the three co-Big 10 champions, MSU, Wisconsin and OSU.  If that had actually been the case, NO WAY does RichRod get replaced.  All this talk on the internet and talk radio and TV isn't even happening.  I'd say that's significant.