Defense Wins Titles… or does It?

One of my biggest pet peeves occurs when an NFL announcer states, as if it’s grade-A proven fact, that defense wins titles. It sounds so simple and catchy, yet when we look at how NFL teams behave, they spend as much on defense as they do on offense. They draft offensive players just as highly as they do defensive players.

Why do NFL franchises, with access to top coaches and scouts, as well as more data than an average fan can even imagine, build rosters in a manner that cannot be defended if the NFL announcer’s simplistic truism has validity?

I studied this several years ago, and came up with data that showed scoring offense correlates about 73% with winning and scoring defense correlates about 70% with winning. Since I’m about to go over my ratings of the all-time best NFL teams, and those ratings depend a lot on that data, I wanted to redo that study.

As with most of my studies, I started with the 1978 season – the beginning of the modern NFL.

First, a look at who reaches the playoffs.

There have been 426 playoff teams in the last 37 seasons. The average playoff team’s offense scores 12.3% more points than the average team in the league. The average playoff team’s defense allows 10.1% less points than the average team in the league.

But does that accurately reflect that the limit on scoring is farther from the average than the limit to a team’s defense? After all, you can only shut out a team, but you can score far more than twice the league average. That could account for the 2.2% difference. In fact, the standard deviation on offense is a little higher (19.0 percentage points) than it is on defense (16.2).

Delving into those 426 teams, 101 scored lower than the league scoring average and another 198 were within one standard deviation over the mean. On the other side of the ball, 97 allowed more points than the league average and another 192 were within one standard deviation over the mean.

No smoking guns here. I really can’t settle the question of offense versus defense based on playoff teams alone. This data suggests that the NFL front office behavior of valuing offense and defense equally is defensible.

But the cliche is, as we know, “defense wins titles.” It’s not “defense wins divisions” or “defense gets you to the conference championship game, but then you’re on your own.”

So I looked at teams that reached the Super Bowl since 1978. That’s only 74 teams.

The average Super Bowl team’s offense scores 25.4% more points than the average team in the league. The average Super Bowl team’s defense allows 16.4% less points than the average team.

The way this difference increases as you reach the Super Bowl does suggest an advantage to the offense. So, again, I looked at the shape of the curve.

Of the 74 Super Bowl teams, one (the 1979 Rams) scored below the league average and 33 scored within one standard deviation of the league average. On defense, eight of the 74 teams allowed more than league average in scoring and 29 more allowed scoring within one standard deviation of the league average. Still a slight edge to the offense.

Trying to break this down further, of the 188 teams that lost their only playoff game, the average offense was 8.2% better than average, with 64 below average and 77 within one standard deviation over the average. The average defense in this group was 8.5% better than average, with 43 below average and another 100 within one standard deviation more than the average.

The next group is teams that won a playoff game, but didn’t reach the Super Bowl. Of those 164 teams, the average offense is 11.1% better than average, with 36 below average and 88 within one standard deviation above the mean. The average defense is 9.0% better than average, with 46 below average and 63 within one standard deviation above the average.

So far, I’m not getting a sense that one side is winning over the other. The data does suggest that mediocre offenses are more quickly eliminated in the playoffs, however.

Ah, but you’re reminding me, “defense wins titles.” It’s not just about reaching the Super Bowl.

Of those 37 Super-Bowl winning teams, 16 were within one standard deviation of the mean on offense and 11 were within two standard deviations. And two were below average on defense (the 2006 Colts and 2011 Giants), 11 within one standard deviation and 16 within two standard deviations.

But then, I looked at actual Super Bowl matchups.

In 21 of the 37 Super Bowls, one team was better both on offense and on defense. That team won 16 of 21 times. It’s those other 16 Super Bowls that caught my attention.

Now 16 is a very small sample size, but the average winning team was 22.8% better than average on offense, and the average losing team was 31.7% better. And the average winning team was 23.3% better than average on defense, but the average losing team was only 10.2% better than average.

What that means is that in the 16 matchups where one team was better on offense and the other better on defense, 12 were won by the better defensive team and only four (including the 2014 Patriots) by the better offensive team.

And in all four cases where the defense was more than one standard deviation better than the other team’s defense while the offense was more than one standard deviation worse, the better defensive team won.

Out of all the data I examined, this was the only piece that stood out for the defense. With everything else, the suggestion that the two are equal or the offense has a slight edge was supported.

As an aside, only in three of the 37 Super Bowls, one team was more than a standard deviation better than the other both on offense and defense. In two of those cases (1985 Bears and 1999 Rams), the better team won. For many reasons, the 2007 Patriots’ Super Bowl loss was the biggest upset in NFL history.

To dig a little bit deeper, I wanted to look specifically at records for teams that were one standard deviation better than average on offense or defense.

On offense, this group consisted of 154 teams (out of 1105). Of those, 128 (83.1%) reached the playoffs, where they had a 152-106 record (.589).

On defense, this group consisted of 163 teams. Of those, 137 (84.0%) reached the playoffs, where they had a 166-113 record (.595).

Thirty-two teams were in both sets. Of those, 30 reached the playoffs and they had a 58-17 playoff record (.773). You’ll find 13 Super Bowl winners and another five Super Bowl losers in that group.

That’s still no reason for the cliche, so I went to 1.5 standard deviations above the mean.

The seven teams that met this criteria both on offense and defense all reached the Super Bowl, with only one team (the 1978 Cowboys) losing – and the 1978 Steelers are also in this elite group of seven. We’re talking about the most dominant teams in NFL history here.

Of the 70 teams that met this criteria on offense, 61 reached the playoffs (87.1%) and earned a 99-44 record (.692).

Of the 59 teams that met this criteria on defense, 57 reached the playoffs (96.6%) and earned an 80-43 record (.650).

You have to go to two standard deviations above the mean to find a big difference here, but only 16 defenses reached that level while 32 offenses reached that level. Those 16 defensive teams were 8-0 in the Super Bowl.

I can’t say that something with that small a sample size is significant. But I’ve searched quite a bit for a smoking gun in the argument that great defenses win Super Bowls, and those eight remaining teams are the argument. The fact that they’re 8-0 rather than 4-4 is essentially why this cliche exists.

Is this random chance? Perhaps. But now you know that when an NFL announcer talks about defenses winning titles, he’s talking specifically about teams that are two standard deviations above the mean on defense, which happens with one team every other year on average and even then he’s only talking about the Super Bowl. That certainly isn’t enough to change the behavior of 32 general managers.

Football Frontier Blog

Welcome to the Football Frontier Blog.

Unfortunately, the old Frontier Blog was lost when our ISP was doing maintenance and deleted the database. While they advertised heavily that they did regular backups of web content, it turns out they don’t.

I was able to find some of the old content in web archives, but only about 20% of the blog. Not enough to restore it in its old form. Many of the entries were more a quick take of then-current events. Those that were the meat of the Blog – the original analysis – could stand to be redone. I have years of new data to work with, and I’m trying to expand and improve my data sets.

The focus of this blog is whatever I see in football that interests me. Primarily, that’s the NFL. But I delve into college football as well. I’ve been ranking college football teams every year since 1997.

As far as pro football goes, I have a much more detailed set of data on games. Play-by-play going back to 1996, though it’s not in any consistent form. One of my projects, should I find the time, is to put these files into a consistent form and draw new analysis.

I also focus a lot on quarterback performance. This database goes back to 1974. I often use 1974 as the beginning of studies because that was the year the NFL opened up the modern passing game by restricting contact with receivers. In 1978, the NFL further restricted contact, and went to a 16-game schedule. Now that we have 37 seasons worth of fairly detailed modern quarterback data, there’s a lot to say about the players who run the show.

This blog has never generated many comments, so I’ve turned off commentary (the spam bots would hit it 100 times a day otherwise). But posts are echoed over at Front Office Football Central and I would encourage (and sometimes join) discussion there.