Does Defense Really Win Championships?

(Photo: r0sss)

The following is a guest post by Tobias J. Moskowitz and L. Jon Wertheimauthors of Scorecasting: The Hidden Influences Behind How Sports Are Played and Games Are Won, now out in paperback. Moskowitz is a University of Chicago financial economist, and Wertheim is senior writer at Sports Illustrated. You may remember Steve Levitt mentioning the book, or the Q&A with the authors on the blog. They are also regular contributors to the “Football Freakonomics” project


Does Defense Really Win Championships?
By Tobias J. Moskowitz and L. Jon Wertheim 

It’s at this point in the NFL postseason when every NFL analyst, pundit, and blogger will inevitably proclaim “defense wins championships.”  With the NFL conference championships upon us this weekend, this phrase will be uttered more times than “yo” in a typical Jersey Shore episode.  And why not?

Last weekend we saw two of the NFL’s top offenses – Green Bay and New Orleans — lose to better defenses.  Moreover, as Chris Berman himself pointed out on ESPN’s Sunday NFL Countdown, 38 (out of 45) Super Bowls have been won by a top 10 defense and 22 have been won by a top three defense. The sentiment has hardened from cliché into an article of sports law. But is it actually true? Does defense really win championships?

In a word: no.

We found that when it comes to winning a title, or winning in sports in general for that matter, offense and defense carry nearly identical weight. For example, here’s what Berman didn’t tell you: the number of Super Bowl champs with a top 10 offense? Thirty-eight. And a top 3 offense? Twenty.  In other words, offense wins championships, too.

We further found that among the 45 NFL Super Bowls, the better defensive team — measured by points allowed that season— has won 29 times.  The better offensive team won 25 times. (Note that adds up to 53, which means that some teams are the better offensive and defensive team in the Super Bowl. Nineteen Super Bowls have featured a team superior on both sides of the ball. Those teams have won 14 of those games.)  It’s a slight edge for defense, but it’s a pretty close call and not different from random chance. The favorite statistic of the “defense wins championships” proponents is that the top-ranked defense during the regular season has won 15 Super Bowls, whereas the top-ranked offense has won only 8. Although this would seem to confer an advantage to defense, these two numbers are not statistically different. And, remember, since the top-three defenses have won no more than the top-three offensive teams, it also means that offensive teams ranked 2 and 3 have won more Super Bowls than the second- and third-best defensive teams, though again, these differences are not statistically significant.

But we’re only talking about 45 games, so let’s broaden the sample size. There have been 427 NFL playoff games over the last 45 seasons. The better defensive teams have won 58 percent of them. The better offensive teams have won 62 percent of the time. (Again, the winning team is sometimes better both offensively and defensively, which explains why the total exceeds 100 percent.) That’s a slight edge to the offense, but again, pretty even.

In almost 10,000 regular season games, the better defensive team has won 66.5 percent of the time compared with 67.4 percent of the time for the better offensive team. That’s a slight nod to the offense but a negligible difference.

But maybe the phrase “defense wins championships” is supposed to mean is that defense is somehow more necessary than offense. Maybe a team can prevail with a middling offense, but not with a middling defense. As it turns out, that doesn’t hold up, either. Three times the Super Bowl champion ranked in the bottom half of the league in defense; only twice did it rank in the bottom half in offense. The lowest-ranked defensive team to win a Super Bowl was the 2006 Indianapolis Colts, rated nineteenth that year. (They offset that by ranking third in offense.) The lowest-ranked offensive team to win the Lombardi Trophy? The 2000 Baltimore Ravens, ranked nineteenth in offense but first in defense.

What about when a great offense faces a great defense?  Twenty-seven Super Bowls have pitted a top 5 offense against a top 5 defense. The best offensive team won 13, and the best defensive team won 14. Another stalemate.

In the NFL it seems, you need either exceptional defense or exceptional offense to win a championship. But neither one is more important than the other.

Okay, but does defense give an underdog more of a chance? Are upsets more likely to be sprung by defensive-minded teams?

Sifting through the numbers, we found that the answer is again no. In the regular season, playoffs, and championships, underdog teams are no more likely to win if they are good defenders than if they are good scorers.

If defense is no more critical to winning than offense is, why does everyone from Little League coaches to ESPN analysts extoll its importance? Well, no one needs to talk up the virtues of scoring. No one needs to create incentives for players to score more touchdowns. There’s a reason why fans exhort “De-fense, De-fense!” not “O-ffense, O-ffense!” Offense is fun. Offense is glamorous. Who gets the Nike shoe contracts and the other endorsements — the players who score or the defensive stoppers?

Quick, which of the following set of names is more recognizable? The top five touchdown leaders in NFL history: Jerry Rice, Emmitt Smith, La-Dainian Tomlinson, Randy Moss, and Terrell Owens? Or the top five interception leaders: Paul Krause, Emlen Tunnell, Rod Woodson, Dick Lane, and Ken Riley?

Bottom line: Defense is no more important than offense. It’s not defense that wins championships. In virtually every sport, you need either a stellar offense or a stellar defense, and having both is even better.

So, who will win this weekend?  Here’s how the teams stack up in terms of offense and defense:

(Note:  every team is really good at either offense or defense.)

It’s a clear offensive vs. defensive matchup in both conference championships.  If you believe the hype, it’ll be a matchup of the defensive-minded Harbaugh brothers in the Super Bowl.  If you follow the numbers, it’ll likely be only one Harbaugh who makes it, but we couldn’t tell you which.


Leave A Comment

Comments are moderated and generally will be posted if they are on-topic and not abusive.



View All Comments »
  1. frankenduf says:

    i like Mark Jackson’s quip, when Kobe scores against a double team: “good defense, better offense”

    Thumb up 2 Thumb down 0
  2. PAT says:

    I’m going to go with the better avg. of offensive/defensive rank in the NFL. Ravens win the Super Bowl. Of the teams remaining, their average off/def rank is 9, where as the Patriots (16.5), 49ers (15), and Giants (17.5) all have poorer averages of the two stats.

    Thumb up 3 Thumb down 6
  3. Matt says:

    The fundamental problem with this analysis is that offense and defense are not orthogonal. Both yards and scoring allowed are somewhat a function of the style and effectiveness of the offense (and the reverse as well). Further, your defense stats ignore forcing turnovers, which often lead to relatively easy scores. There are many offensive and defensive effectiveness metrics (e.g., football outsiders) which provide much more sophisticated measures.

    I am not suggesting the conclusions here are necessarily incorrect, but I am not sure variables are sufficient for the analyses you are conducting.

    Well-loved. Like or Dislike: Thumb up 21 Thumb down 2
    • Bill Bowen says:

      Matt’s assessment of the varibales is absolutely correct. I can’t dispute whether the conclusion of the article is correct, but the methods it uses to draw it’s conclusion are deeply flawed.

      To say, for example, that New England’s defense was the penultimate defense is the NFL (and I’m no fan of N.E.) is based solely on yards and points allowed. But these two outdated gauges are brutally unfair to defenses like New England’s, Atlanta’s, New Orleans or any team that employs a no-huddle offense. Those teams offenses get their opponents down so far so fast, that the other team is forced to go to no huddle offenses as early as the 2nd quarter just to try to stay in the game. This keeps the good offensive team’s defense under constant assault, and directly contributes to more yards allowed and more points allowed.

      A better tool for measuring defense is the “Bendability Index”. Divide the defenses yards allowed by the points they allowed (which gives you points per yards allowed), and you see a very different picture of which defenses “bend”, but don’t break. Defenses like Houston and Seattle still appear at the top, but you’ll see San Francisco and New England’s defense right behind them.

      Thumb up 1 Thumb down 1
  4. Tony says:

    That image is labeled “defense,” but I’m pretty sure that’s a picture of the third, less-discussed-but-equally-important, phase of the game: special teams. (Specifically, it looks like the Patriots are in kick return formation.)

    Well-loved. Like or Dislike: Thumb up 12 Thumb down 1
  5. tony says:

    This analysis is very oriented toward “rankings” of O and D.

    Until I see an explanation of how “rankings” are determined… and the conditional variance of such, I’m unwilling to accept the conclusion of the article.

    Consider a basic football strategy:

    -(if you get) a lead

    - grind out the clock by running the ball (produces fewer yards, but chews up time)

    -win by a small margin

    This is a strategy that (is believed to) maximize the chance of winning the game… but may have a negative effect on the “ranking” of your offense.

    I also can’t really believe any football analysis that doesn’t even really mention “special teams”. I assume that field goal kicking is lumped in with offense… is punting included in defense ranking?

    Well-loved. Like or Dislike: Thumb up 8 Thumb down 1
    • Tony says:

      Rankings are based on yards allowed. The defense that allows the least yards ranks number one, regardless of points scored against or winning percentage. The offense that gains the most yards is ranked number one, regardless of points scored or winning percentage.

      Thumb up 1 Thumb down 0
      • Tony W says:

        There you have it… turn the ball over in your red zone enough and your defensive ranking will be stellar!

        Thumb up 1 Thumb down 0
      • Matt says:

        Right. The assumption is that yards gained/allowed is a better proxy of how effective an offense/defense is, because it doesn’t include all the flukey ways a team can score. This is the norm when talking about “Top Offenses” or “Top Defenses.”

        Thumb up 0 Thumb down 0
      • MikeM says:

        Actually, no. He says at the outset that defense in this case is measured by points allowed over the season.

        So in that case, I don’t really trust the data. Remember: in football, defenses score points too.

        Thumb up 1 Thumb down 1
  6. Michael says:

    this is one of my favorite misunderstood quotes. Too many geniuses hear this saying and think you don’t need an offense anymore, or that you can get away with a faulty one. The saying is Defense wins CHAMPIONSHIPS. It doesn’t say regular season games, or will take you to a championship, it says defense wins championships. It can be assumed that if a team is in the superbowl that they know how to score points. When an intelligent person is saying it, they mean that when 2 good teams who can score proficiently come together, the better defense makes the difference. Anyone who thinks more or less than that hasn’t thought it through.

    Thumb up 4 Thumb down 7
  7. emily says:

    Has anyone thought about normalizing “total yards” for offense and defense by time of possession? That is, computing average rates of progress per unit time when the defense and offense are on the field. It seems like a defense that allows 200 yards in a game where the other team possessed the ball 75% of the time is doing a better job than a team that allows 100 yards in a game where the other team possessed the ball 25% of the time. Same story for offense. I don’t know that much about American football- is the possession breakdown usually close to 50/50 or are there pretty big differences there that should be accounted for?

    Thumb up 2 Thumb down 0
    • MikeM says:

      Actually, that’s a nice idea but I don’t think it would be as telling as you’d want it to be. Some teams utilize running offenses better, or defense the run better than passing, which can skew time of possession rather heavily.

      One interesting measure of offensive and defensive prowess might be a curve charting points-scored as a function of starting field position. The x-axis would be the yard markers of the field, and the y-axis would be points-scored. Obviously, as you approach the goal line, the curve would approach something just short of 7 (if you count extras point tries, which I would).

      This could work for both offense and defense and a quick glance at the chart would give a knowledgeable person a real sense of the strength of the team. Even better would be a concurrent histogram of the frequency of possessions at each field position.

      Thumb up 0 Thumb down 0
      • MikeM says:

        RE: ABOVE

        By “points-scored” I mean either “points-scored” or “points-allowed” depending on whether you’re measuring offense or defense.

        Thumb up 0 Thumb down 0
  8. RGJ says:

    My grips with stuff like this is the same with fantasy football — too often 4th quarter stats in games well in hand skew the “true” match of competitive games.

    Case in point:

    New England leads the Colts 31-3 after 3 quarters in Week 13. Colts score 21 points ho-hum prevent defense points.

    The other aspect is the incredible irrelevance of season long stats in predicting the next game. The Giants are the 27th ranked defense, but after a year of injured defense stars, they have now given up an average of only 11 points a game over the past 4 weeks in beating three top ten offenses. They are en fuego on defense despite their 27th ranking.

    There are just way too many factors for blanket theories in football — injuries, home field, weather, recent trends, special teams.

    BTW, Giants 28, SF 17 :-)

    Thumb up 4 Thumb down 2
    • ryan says:

      Why are you using inferential tests when you have access to population data? To which other NFL are you generalizing your results? Statistical significance is meaningless when you have access to all the entire universe of cases.

      Also, why would you deem 45 to be too small a sample?

      Thumb up 1 Thumb down 2