# Tempo Plus Turnovers Help To Explain Offense and Defense

Submitted by Enjoy Life on November 17th, 2011 at 11:48 AM

Synopsis: There have been literally thousands of posts about the decline in Michigan's offense and the improvement of the defense this year. After 10 games, some of the raw numbers are staggering. Michigan's defense has allowed 52% fewer points per game and the offense is scoring 14% fewer points per game. There were also literally thousands of posts debating what effect the tempo offense had on both offense and defense. If we had tried to create an experiment to answer these questions, we probably could not have done as well as the comparison we now have between 2011 and 2010.

To provide valid comparisons, it is necessary to make two adjustments; (1) Delete the impact of the OT game in 2010 – both points and plays, and (2) adjust for turnovers in 2011 and 2010. After these adjustments, we can incorporate tempo by calculating points per play. Here are the results.

As you can see, after adjusting for tempo and turnovers, Michigan's offense is actually scoring more points per play this year as compared to last year. And, after those same adjustments, about 55% of the improvement in defense can be attributed to tempo and turnovers. Of that 55%, approximately 18% is due to tempo and 37% is due to turnovers.

Michigan's offense has had 12% fewer plays this year and the defense has had been on the field for 16% fewer plays. This is primarily due to the tempo of the offense. There can be no doubt that the tempo of the offense was a significant reason for poorer performance of the defense in 2010.

BTW, turnovers are NOT primarily a matter of just luck but are a result of better performance. Better performing offenses will have fewer turnovers, better performing defenses will force more turnovers. And, these turnovers help make the offense and defense even better.

A Look At The Numbers: A direct comparison of raw data from 2011 and 2010 is simply not valid. The most common error is a failure to recognize the impact of a 3 OT game in 2010. This added 22 points to the offense and 20 points against the defense. This is approximately a 6% increase.

In order to determine the impact of turnovers, I already had the numbers from 2011 because of my weekly turnover analysis. But, I had to go back to 2010 and calculate all the expected points for each turnover of each game. This was not a pleasant experience as I had to relive the catastrophe of 27 giveaways and only 18 takeaways (over 10 games).

Each turnover usually results in a loss of expected points to the team losing the TO and, usually a gain in expected points to the team gaining the TO. In most cases, the loss in expected points is greater than the gain. Unless a team has no turnovers, the net result is that both teams lose points. The team with fewer turnovers simply loses fewer points.

In 2010 Michigan's offense lost 22 turnovers thru 10 games and the defense had only 15 takeaways. Without the TOs, M would have scored 31 more points and the opponents would have scored 18 more points.

This year M has lost 19 turnovers but the defense has had 23 takeaways. Without the TOs, M would have scored 15 more points this year and the opponents would have scored 50 more points.

Why The Difference In Expected Points?: The expected points for each TO are calculated based on the down, spot TO was lost, and spot TO was gained. The maximum value for a turnover is 11.7 EP (1-10 at the 1 yard line and the TO is returned for a TD), and the minimum is –1.2 (4th down on the 40 and the pass is intercepted on the 1 yard line – ball should have been knocked down and team would have taken over at the 40). This year, M had a TO worth 11.03 EP (Heron's return from the M04 for a TD) and another TO worth –0.53 (interception by EMU on a 4-5 from the V34 that resulted in EMU getting the ball at the V27).

In addition to the turnover margin, several other reasons the EP are significantly different between 2011 and 2010 are:
1. M has 3 returns for TD this year – none in 2010.
2. Opponents have lost 6 TOs in the red zone this year – just 3 in 2010.
3. M has 5 fumbles lost and 14 interceptions this year – 10 fumbles lost and 12 interceptions in 2010 (fumbles have greater impact).
4. Opponents have lost 16 fumbles and 7 interceptions this year – 6 fumbles and 9 interceptions in 2010 (fumbles have greater impact).

Do you have the numbers just for B1G games?  It wouldn't surprise me if the improvement there was even greater. I might want to back out the Illinois game or see the data  with and without that game.. I'll take 45 fewer points on O against opponents scoring 114 points fewer any day of the week, particularly when you are still averaging 34 points per game on O.

For 2011 stats, did you include WMU for completeness' sake?

Yes, I included WMU (used MGoBlue.com data)

I completely disagree with the NCAA about the WMU game. The distortion caused by not including the game is obviously greater than the distortion of including the game. Especially since several NCAA rankings are based on absolute numbers and not on per game numbers.

be 9.75 instead of 10 for 2011 due to the missing 4th quarter?

On the per game stats, perhaps we should use 9.75 instead of 10. Of course, that is another reason the per play stats make more sense. Less distortion.

The other question would be whether M and WMU would have scored any more points under those conditions.

There can be no doubt that the tempo of the offense was a significant reason for poorer performance of the defense in 2010.

1. You ignore special teams totally. Last year we had 4 FGs all season. This year, we have 8 already. That's 12 more points that need to be controlled for. Removing 4 of those to put us on pace with last year gives PPP of 0.508. Otherwise you're attributing the scoring to the offense, which isn't the case.

2. I concede that deleting OT is valid, in that playing with the short field skews the scoring numbers. However, deleting the entire game should not be valid. The first 4 quarters of the game should stand on there own.

3. The season ain't over. If you're want to compare apples to apples, compare 2010 through Purdue to 2011 through Illinois. Also, last year's SOS was higher, so that level of opposition should be controlled for as well.

BTW, turnovers are NOT primarily a matter of just luck but are a result of better performance. Better performing offenses will have fewer turnovers, better performing defenses will force more turnovers.

Here's the turnover rank for the top 20 teams scoring defence from the NCAA:

 TO rank name 59 TCU 26 Boise St. 30 Alabama 49 West Virginia 12 Ohio St. 18 Missouri 45 Iowa 49 UCF 49 Nebraska 12 Stanford 8 LSU 2 Oregon 72 Clemson 36 Northern Ill. 59 Pittsburgh 72 Temple 94 Syracuse 77 Louisville 7 Boston College

No correlation. Ditto for offense (I'll save you the chart). Look at the numbers. Out of 600 to 700 plays, 30 TOs is about 4%. Even if it was 10 TO, it's still 1.5%. It's just too small a factor to matter.

I'm pretty sure he deleted just the OT, not the whole game.  Can we get a clarification on that?

Yes, of course, it was just the Illini OT. Notice in the table that I just deleted 20/22 points and 14/11 plays.

Where is MSU? USA Today has them at 8th in the nation in scoring defense as of Nov 13 @ 9:42am

You are correct, that isn't the top 20 scoring defenses and I don't know what that list is.

Top 20 scoring defenses and turnover rank:

 1 Alabama 78 2 LSU 17 3 Penn St. 17 4 Temple 85 5 Wisconsin 93 6 Florida St. 50 7 Michigan 24 8 Michigan St. 30 8 Virginia Tech 85 10 UCF 105 11 Rutgers 7 12 Georgia 17 12 Illinois 39 12 Louisville 105 15 Boise St. 24 16 Ohio St. 78 17 Mississippi St. 62 17 Utah 8 19 South Carolina 3 20 Florida 111

Boise State has only played 9 games while everyone else on this list has played 10.

Turnover ranking link is below.  Couldn't find a site that just ranked takeaways as they all sorted by turnover margin.
http://www.teamrankings.com/college-football/stat/takeaways-per-game

Not sure I understand a lot of this comment. But, here goes.

(1) All data is thru the first 10 games (this year and 2010). Last 2 games in 2010 were Wisc and osu. Last 2 games this year are Neb and osu. Data seems as comparable as you are likely to get.

(2) So, you are saying that the fact that M's opponents had 16% fewer plays thru 10 games  had no impact on the points they scored? That seems unrealistic. Therefore, I concluded that tempo had an impact.

(3) I'll take another look at the data correlation (see 3a below). That said, if you belive TOs are primarily luck then you must also belive that RR was about the most unlucky coach on the planet (3 straight years of double digit negative turnover margin) and Brady Hoke is just getting lucky this year. I just find that hard to believe.

(3a) Here's a great site for TO analysis:

I have been looking mainly at won-loss record versus turnover margin. In a previous diary

there was a good correlation. I'll have to look closer at the defensive correlation over the past several years.