Downs And Success And Differentials: A Further Look

Submitted by LSAClassOf2000 on August 7th, 2013 at 9:28 AM

DOWNS AND SUCCESS: A FURTHER LOOK            

After not finding as much as I thought I would in the initial delve into this subject, a few suggestions from other MGoBlogger and some folks on Twitter proved to be rather enlightening (thanks to Tenn Blue and jamiemac in particular). I was taking an incomplete approach to this subject, and in some respects, one that might be a little too broad to find much.

So, I limited this to 3rddowns because that is where some of the better relationships seem to be, and then I calculated both the conversion rate on offense and the conversion percentage on defense for each team. Next, I subtracted the defensive success rate from the offensive success rate to get a differential for each team and each year evaluated. I also took the suggestion to correlate this to certain aspects of scoring offense – for this run, I chose average points per game.

The R-value for the relationship between the differential and win percentage turns out to be 0.752, which is not bad at all. The R-value for the relationship between the same differential and average points per game came out to 0.659, so while not as good, the two share at least a casual relationship, as you will be able to see in the graphs below.

A FEW HIGHLIGHTS REGARDING THIRD DOWNS:

So, what did some of the better teams from the Big Ten (I included Rutgers and Maryland in here for kicks) look like when it came to success on getting conversions, stopping them on defense and winning?

The best ten from the Big Ten data set:

Year

Name

DEFENSE PCT

OFFENSE PCT

DIFFERENTIAL (OFFENSE-DEFENSE)

Win. Pct

2008

Penn St.

32.26%

52.02%

19.77%

0.846

2005

Ohio St.

30.86%

49.36%

18.50%

0.833

2009

Penn St.

31.77%

48.35%

16.58%

0.846

2011

Wisconsin

38.22%

54.71%

16.49%

0.786

2006

Ohio St.

32.95%

49.06%

16.10%

0.923

2010

Ohio St.

30.18%

46.20%

16.02%

0.923

2012

Michigan

35.96%

50.29%

14.33%

0.615

2008

Nebraska

33.73%

47.19%

13.46%

0.692

2006

Wisconsin

28.18%

41.42%

13.24%

0.923

2007

Ohio St.

33.33%

45.90%

12.57%

0.846

 

And to follow, the worst ten:

Year

Name

DEFENSE PCT

OFFENSE PCT

DIFFERENTIAL (OFFENSE-DEFENSE)

Win. Pct

2008

Indiana

46.78%

27.65%

-19.14%

0.250

2011

Maryland

53.76%

35.84%

-17.93%

0.167

2005

Illinois

50.67%

37.57%

-13.09%

0.182

2006

Minnesota

48.81%

36.75%

-12.06%

0.462

2008

Michigan

38.78%

27.27%

-11.50%

0.250

2010

Purdue

43.03%

32.22%

-10.81%

0.333

2009

Minnesota

44.39%

34.08%

-10.31%

0.462

2011

Indiana

49.39%

40.82%

-8.57%

0.083

2009

Indiana

47.50%

39.08%

-8.42%

0.333

2006

Indiana

41.94%

33.53%

-8.40%

0.417

 

You’ll note that Michigan has one representative in each of these tables of extremes.

Here is a general table which shows you what the average differential was in various ranges of win percentage.

 

WIN PCT.

AVG. DIFFERENTIAL

0.800 OR BETTER

11.60%

0.799 TO 0.700

8.36%

0.699 TO 0.600

5.11%

0.599 TO 0.500

2.49%

0.499 TO 0.400

-2.03%

0.399 TO O.300

-3.86%

0.299 OR WORSE

-7.86%

 

THE POINTS PER GAME ANGLE:

I actually thought about going back and getting data on field goal attempts as well for this one, but I wanted to test out the relationship between this differential I was working with and the most basic statistic for scoring offense first. First, the 10,000 foot view of the relationship:

 

POINTS PER GAME

AVG. DIFFERENTIAL

35 OR MORE

13.45%

34-30

5.78%

29-25

3.96%

24-20

-2.48%

LESS THAN 20

-6.01%

 

FOR EACH TEAM:

Below are the comparisons of the 3rddown differential to both win percentage and points per game for each team in the Big Ten, including our soon-to-be conference compatriots.

 

 photo Wisconsin3rdDownWin_zps8de42353.png  photo Wisconsin3rdDownPPG_zpse9475066.png  photo Rutgers3rdDownWin_zps7c6b117a.png  photo Rutgers3rdDownPPG_zps3e0e6bad.png  photo Purdue3rdDownWin_zpsadf83511.png  photo Purdue3rdDownPPG_zps3f2a13ea.png  photo PSU3rdDownWin_zps376356d2.png  photo PSU3rdDownPPG_zpsc009f911.png  photo OSU3rdDownWin_zpsf85d3d3f.png  photo OSU3rdDownPPG_zps8c95fba1.png  photo NW3rdDownWin_zps3e09ba90.png  photo NW3rdDownPPG_zps10c48819.png  photo Nebraska3rdDownWin_zps1e20b875.png  photo Nebraska3rdDownPPG_zpsa62b9e9e.png  photo MSU3rdDownWin_zpsee0ac411.png  photo MSU3rdDownPPG_zps324a9851.png  photo Michigan3rdDownWin_zpsab714892.png  photo Michigan3rdDownPPG_zps01b913df.png  photo Maryland3rdDownWin_zpseb21ee05.png  photo Maryland3rdDownPPG_zps66b12482.png  photo Iowa3rdDownWin_zps5a8241cb.png  photo Iowa3rdDownPPG_zpsf1cc2973.png  photo Indiana3rdDownWin_zps9d0a914e.png  photo Indiana3rdDownPPG_zpsfcb46e13.png  photo Illinois3rdDownWin_zps40046d7f.png  photo Illinois3rdDownPPG_zpsae0351d0.png

 photo Minnesota3rdDownWin_zps85a9a44c.png  photo Minnesota3rdDownPPG_zps85c46fa2.png

CONCLUSION:

I was a little happier with the results of this, even though it again confirms something that you already thought about the game - the fewer 3rd downs that you convert on offense and the more that you allow your opponent to convert, the less you score and the worse you will do overall, at least most of the time. Like any sport, there are several things that confound this and created exceptions, such as teams that have good defenses but also middling offenses (like Michigan State).

OBLIGATORY:

 

Comments

Daniel

August 7th, 2013 at 10:46 AM ^

what the differential would predict: OSU 2012 is really high, and Michigan 2012 is really low.

I'm not sure exactly what it means for OSU. Either they got really lucky last year, or Urban Meyer is a really fantastic coach? I'm leaning toward a serious dose of the former and a middling dose of the latter.

For Michigan, I would assume it's mostly luck and facing difficult opponents.

Edit: also, poor Illinois and Michigan State 2012

stbowie

August 8th, 2013 at 12:58 PM ^

It's unclear (except from context clues) whether the defensive percentages are opponent's success rates on completing third downs, or defensive success rates for getting stops. I assume it's the former, but thought I'd suggest making that explicit. Nice work here!