How to Win a Super Bowl

How to Win a Super Bowl

This is an attempt to determine and put in order what statistical categories most strongly correlate with Super Bowl success. Because I am not manipulating any variables, it is not statistically sound to say what categories actually “cause” a team to win a Super Bowl. What we can do, however, is look at what Super Bowl winning teams do most differently than the competition and what they do best internally. Do they pass the ball more or less than other teams? Are they better at running the ball or stopping the run? Do more teams have high flying offenses or shut down defenses?

Deciding what statistical tests and values to use to compare the different variables is not necessarily straightforward. Doing a simple comparison of averages is not ideal because there is so much variance among the teams. Stopping the run is one of the most important qualities of a winning team, but the 2006 Super Bowl winning Indianapolis Colts, for instance, were actually the worst in the league at it that season, giving up an absolutely ludicrous 173 rushing yards per game. Furthermore, many statistical categories trend in specific directions through the course of NFL history. So how do we say which values truly stand out? We calculate T-values. T values show the true difference between sets of averages. The larger the magnitude (Read: amount from 0, so a T value of -3 will be “better” than a T value of +2) of the T-value is for any category, the more the category stands out from the pack. [1]

There is an important caveat, however. We should not confuse correlation with causation. For instance, out of all the categories I looked at, offensive passing yards per game ranked towards the bottom, meaning that Super Bowl winning teams don’t pass the ball for that many more yards than the rest of the league overall. Does this mean that a moderate passing attack is better suited for victory than a strong one? Probably not. It likely has to do more with the fact that winning teams get out to leads early and then spend the later parts of games running the ball instead of passing. 

So here is what we are looking at, from 1978 (the start of the 16 game season) to 2017. 

Offensive categories: Points per game, total yards per game, pass attempts per game, pass yards per game, rush attempts per game, rush yards per game, total turnovers, total sacks allowed.  Defensive categories: Points per game, total yards per game, pass yards per game, rush yards per game, total sacks, total forced turnovers. 

So first I have graphs of these categories, ordered by magnitude of T value. The left column is the team that won the Super Bowl that year. To the immediate right, is the league average that year. The numbers at the top are where the Super Bowl winning team ranked each year in that category. The text under each graph shows the T-value, the average numbers for the super bowl winning teams + the average rank, the average of every other team, the winning team with the highest numbers + their rank, and the team with the lowest numbers + their rank. After the charts is some analysis.

Lastly are some random observations I made while compiling the data. 

Here is the Excel spreadsheet I put together. It’s got everything on the charts plus lots of fun stuff on it that didn’t make it to the blog like Pro Bowl players per team, average age, and more.


A couple of quick notes: 

In 1982 and 1987 there was a player strike in the NFL. The season was shortened to 9 games in 1982 and in 1987, scab players were brought on as replacements for part of the season. Because of this, I left out a lot of those numbers from the charts and calculations. Lastly, for the life of me, I could not get the charts to embed properly. Therefor, they are not interactive like I would prefer them to be, just pictures that cut out at the bottom. Sorry. You can still click on them to enlarge.

I got most of the data from Profootballreference.com and a bit from wikipedia and nfl.com. The graphs were made with a handy site called plotly (plot.ly.) Image credit goes to Sportingnews.com

Offensive Points Game

Offensive Points Per Game:

T stat: 8.09

Winner mean/rank: 25.72 (6.35)

League wide mean: 21.06

Highest winner/rank: 32.88 (1999 St louis rams) (1)

Lowest winner/rank: 20.8 (2000 Baltimore Ravens) (14)

Points Per Game Allowed

T stat: -8.07

Winner mean/rank: 16.67 (5.58)

League wide mean: 21.06

Highest winner/rank: 10.31 (2001 Baltimore Ravens) (1)

Lowest winner/rank: 25 (2011 New York Giants)

Total Yards Per Game

T stat: 5.52

Winner mean/rank: 353.41 (8.69)

League wide mean: 325.6175

Best winner/rank: 403.81 (2009 New Orleans Saints) (1)

Worst winner/rank: 300.31 (1990 NY Giants) (17)

Total Yards Allowed Per Game

T stat: -5.27

Winner mean/rank: 296.39 (7.55)

League wide mean: 325.62

Best winner/rank: 237.19 (2008 Pit Steelers) (1)

Worst winner/rank: 376.38 (2011 NYG) (27)

Rush Yards Allowed Per Game

T stat: -5.19

Winner mean/rank: 98.69 (7.49)

League wide mean: 116.35

Best winner/rank: 60.63 (2000 Baltimore Ravens) (1)

WORST winner/rank: 173 (2006 Colts) (32)

Total Sacks

T stat: 4.22

Winner mean/rank: 43.93 (9.10)

League wide mean: 37.63

Highest winner/rank: 64 (1985 Chicago bears) (3)

Lowest winner/rank: 25 (2006 Indianapolis colts) (25)

Rush Attempts Per Game

T stat: 4.12  Winner mean/rank: 31.45 (8.33)  League wide mean: 28.76  Highest winner/rank: 40.06 (1978 Pit)(3)  Lowest winner/rank: 25.69 (2011 NYG) (22)

T stat: 4.12

Winner mean/rank: 31.45 (8.33)

League wide mean: 28.76

Highest winner/rank: 40.06 (1978 Pit)(3)

Lowest winner/rank: 25.69 (2011 NYG) (22)

Rush Yards Per Game

T stat: 3.88

Winner mean/rank: 128.83 (9.93)

League wide mean 116.35

Highest winner/rank: 172.56 (1985 bears)

Lowest winner/rank: 89.19 (2011 NYG)(32)

Total Forced Turnovers

T stat: 3.31

Winner mean/rank: 35.45 (8.82)

League wide mean: (30.38)

Highest winner/rank: 54 (1985 chicago bears)(1)

Lowest winner/rank: 23 (2016 patriots)(14)

Total Turnovers Allowed

T stat: -2.76

Winner mean/rank: 25.84 (8.95)

League wide mean: 30.38

Best winner/rank: 11 (2016 NE Patriots)(1)

Worst winner/rank: 52 (1979 Steelers) (28)

Pass Yards Per Game

T stat: 2.76

Winner mean/rank: 224.56

League wide mean: 209.27

Highest winner/rank: 295.88 (2011 NYG) (5)

Lowest winner/rank: 168.69 (1978 steelers) (12)

Total Sacks Allowed

T stat: -2.53

Winner mean/rank: 33.25 (13.2)

League wide mean: 37.63

Best winner/rank: 9 (1991 Redskins)

Worst winner/rank: 55 (1983 LA raiders)

Pass Attempts Per Game

T stat: -1.67

Winner mean/rank: 31.51 (17.95)

League wide mean: 32.54

Highest winner/rank: 38.06 (2014 New England Patriots)

Lowest winner/rank: 23.69 (32)

Pass Yards Allowed Per Game

T stat: -1.43

Winner mean/rank: 202.16 (11.7)

Average mean: 209.27

Best winner/rank: 149.63 (1978 Steelers) (12)

Worst winner/rank: 269.25 (2006 Colts)

Takeaways 

So here is what we have, in order of T value magnitude. 

PPG O , PPG D , YPG O , YPG D , Rush Yards D , Sacks D , Rush Attempts O , Rush Yards O , Forced Turnovers D , Turnovers O , Pass yards O , Sacks allowed , Pass attempts O , Pass Yards D. 

The 4 most important stats based on the T values are the major offensive and defensive categories, points and yards. This is not much of a surprise.  The T-value of points for ( 8.09) just barely beats out points against (-8.07), but it is worth noting that, on average, the points against ranking (5.58) beats out the points for (6.35) ranking by almost a full spot. As John Madden so thoughtfully pointed out, “Usually the team with the most points wins the game.” Next, almost 2.5 Ts after, we have total offensive yards shortly followed by total yards allowed. These follow the same trend as the points, the offensive category has a slightly smaller T value magnitude ( -5.27 vs 5.52) but a better overall rank (7.55 vs 8.69.) So, the teams are pretty even on offense and defense points wise, and are also pretty even on offense and defense yards wise. However, there is a pretty big jump between the points and yardage categories. This suggests that good teams, at least defensively, might “bend but not break.” They give up yards, but make up for it by stopping the opposition from actually getting in the endzone. 

There is an idea that a team needs to stop the run to have post season success and this is supported by the data, at least at a correlation level. The Super Bowl winning teams stop the run very well (rank=7.49, T stat=-5.19, avg=98.69 YPG) compared to the rest of the league (avg =116.35 YPG.) When we compare the winners rush defense to their pass defense though, we get our first surprise. Not only are they much better at rush defense then pass defense (T=-1.43, rank= 11.7), they don’t actually allow that much less passing yards than the rest of the league (202.16 vs 209.27.) In fact, pass yards against actually has the smallest T-value magnitude out of everything I looked out. This probably has a lot to do with opposing teams being down and turning primarily to passing as the game progresses, however. Offensive rushing vs offensive passing follows a similar trend and can likely be chalked up to the same cause, as the rush numbers of winning teams stand out way more than the passing numbers:

Rush Attempts O 

T stat:  4.12

Winner mean/rank: 31.45 (8.33)

Average mean: 28.76

Rush Yards O 

T stat: 3.88

Winner mean/rank: 128.83 (9.93)

Average mean: 116.35

Pass yards O 

T stat: 2.76

Winner mean/rank: 224.56

Average mean: 209.27

Pass attempts O 

T stat: -1.67

Winner mean/rank: 31.51 (17.95)

Average mean: 32.54

So Super Bowl winning teams are more effective at running the ball than passing, and they are more effective at stopping the run than stopping the pass, but take it with a grain of salt. 

The next thing that stands out is how much better the Super Bowl winning teams seem to be at getting after the opposing quarterback than protecting their own. The T stat is better for defensive sacks vs sacks allowed (4.22 vs -2.53) as well is the average rank (9.10 vs 13.2.)

Lastly, we have turnovers for vs turnovers against. Once again, defense prevails. Super bowl winning teams stand out more with forcing turnovers (T stat=3.31, rank=8.82) than they do with keeping the ball safe (T stat=-2.76, rank = 8.95)

Stray Observations

The 2006 Colts Defense was Horrific

The first Super Bowl that I remember vividly was the 2007 matchup between the Colts and the Bears. The two teams were polar opposites. The Bears defense was one of the best in the league and featured the likes of Brian Urlacher, Peanut Tillman, and Lance Briggs. Their offense had…..Rex Grossman. I actually remember rooting for the Bears that game because I felt so bad for Grossman, who seemed like a great guy despite being a horrific quarterback. The Colts, on the other hand, were an offensive team through and through. They had Peyton Manning in his prime throwing to 2 of the greatest receivers of all time, Reggie Wayne and Marvin Harrison who both put up unreal 1300+ yard seasons. They even had a respectable rushing game, led by 1000 yard back Joseph Addai. Combing through all the numbers though, what stood out to me was not how fantastic the Colts offense was that season, but what an unmitigated dumpster fire their defense was. They ranked 23rd in points per game allowed, 30th in rush touchdowns allowed per game, 32nd in TOP allowed. Remember David Carr? He threw for 3 touchdowns against them and had an 140.2% QB rating. Bad I know? It gets worse. THEY GAVE UP A LEAGUE WORST 173 RUSHING YARDS PER GAME. So, defense wins championships, unless you have Peyton Manning I guess. 



2000 Ravens D may have Been Better than the ‘85 Bears

Everyone knows that the 85’ bears had the best defense of all time. But did they actually? Take a look at the defensive numbers. 

Points Per Game

2000 Baltimore Ravens: 10.31

1985 Chicago Bears:12.38

Total Yards Per Game

2000 Baltimore Ravens: 247.95

1985 Chicago Bears: 258.44

Pass Yards Per Game

2000 Baltimore Ravens: 187.31

1985 Chicago Bears: 176

Rush Yards Per game

2000 Baltimore Ravens: 60.63

1985 Chicago Bears: 82.44

Total Forced Turnovers

2000 Baltimore Ravens: 49

1985 Chicago Bears: 55 

Total Sacks

2000 Baltimore Ravens: 35

1985 Chicago Bears: 64

I think given that the Ravens were playing in a more offensively geared era, you actually have to give the greatest defense of all time title to them. Sorry Mike Ditka. 



Perhaps I am Biased but the 2011 and 2007 giants had to be two of the Most Garbage TeamS to Win a Super Bowl. 

So Im a strong believer in that the team that wins the Super Bowl is the best team and vice versa. This might seem obvious but some people would argue that the best team and the team that wins the Super Bowl can be two separate things. I disagree. Buttttttt, If there was a case where the team that won wasn’t actually the best team, or even kind of sucked, it would both the 2007 and 2011 Giants. I’m a Cowboys fan but the numbers don’t lie. They both kind of sucked. Just take a look at the numbers here: 

The NFC East Absolutely Dominated in the 80s-90s.  

The Cowboys were not the only dominant NFC East team in the late 80’s through the mid 90’s. From 1986-1995, 7/10 Super Bowl winners came from the NFC East. The Giants won in 86 and 90, the Redskins in 87 and 91, and the Cowboys in 92, 93, and 95. The 49ers conveniently filled in the gap the rest of those years, winning the Super Bowl in 88, 89, and 94. 

Poor Redskins 

So the redskins have been the picture of mediocrity for the last couple of decades, but in the 80’s and early 90’s they were actually very formidable. In some cruel twist of fate though, two of their three Super Bowls happened in contentious seasons, which I can only assume took some of the spotlight off of them. In 1982 a players strike shortened the season to 9 games and in 1987, replacement players were used for part of the season. Poor Redskins.

Does Having an 1000 Yard Rusher or Receiver Matter?

1000 yards is generally considered the benchmark for a “good” season as a running back or wide receiver. So does having an 1000 yard player seem to correlate with winning the super bowl? 23 out of the 37 super bowl winners (leaving out the 1982 and 1987 seasons) had an 1000 yard rusher. 25 had at least one 1000 yard receiver, and 8 had 2. Seems important, or at least strongly correlated. If you want to see all the numbers, click the fun spreadsheet button up top.

Bonus Spreadsheet: Whats up with the Running Back by Committee Approach/ are NFL Teams Actually Running the Ball Less?

There is a common notion that the running back by committee approach, that is spreading carries around between two or more running backs, has supplanted the lead back strategy in the NFL. Is this actually the case? From 1980-2017, 17 out of the 38 teams had a single back with at least 50% of the teams total rushing yards. It turns out though, that the slope is just a minuscule 0.000350148%, meaning that the percentage of lead back yardage has indeed trended downwards, but just barely. These numbers are just for the super bowl wining teams, but they have not really changed much for them. There is a lot more to unpack here, so take a look at all the numbers here: 

Terrell Davis Was Nasty

Terrell Davis was an afterthought coming out of Georgia in 1995, getting picked by the Broncos in the 6th round. After a great rookie season, he put together what might be the greatest three year stretches by a running back of all time (courtesy of NFL.com.)

1996: 1,538 yards, 13 touchdowns, Offensive Player of the year
1997: 1,750 yards, 15 touchdowns, Super Bowl MVP
1998; 2,008 yards, 21 touchdowns, NFL MVP, Super Bowl champion

And those are just his regular season numbers. He averaged 143 rushing yards per playoff game, a monster 5.59 yards per carry, and is tied for the most 100 yard rushing games of all time in the playoffs, hitting the mark in 7 out of 8 games.

Then it was over. In the following 3 injury riddled seasons, the last of his career, he only played in a total of 17 games and gained less than 1200 total rushing yards. One can only wonder the records TD would have broken if he hadn’t been so injury riddled. Thankfully, he finally snuck into the Hall of Fame in 2017 after being on the ballot for 9 years.

Totally not desperate or anything but if you enjoyed this post please subscribe to my email list way down below.

[1] It is important to realize that T values don’t actually mean much on their own. Comparatively, they show you how different 2 averages are from another. To determine weather this difference is significant or not, you need to calculate something else, a P value. I’ll get deeper into this another time.

Books That Have had a Profound Impact on me

Books That Have had a Profound Impact on me