A new football season is under way, and it won’t be long before someone, somewhere makes the claim that “a good defense beats a good offense” or, more broadly, that “defense wins championships.”
You’re bound to hear this claim made again and again, and many people apparently believe it. But does it hold up to statistical scrutiny?
In the current issue of Chance, statistician Scott M. Berry of College Station, Texas, describes his effort to investigate this claim. He concludes that, in general, the quality of the offense matters more than the quality of the defense in determining a game’s outcome.
To test the claim, Berry developed an equation to model the quality of National Football League (NFL) teams, taking into account home field advantage, offensive prowess, and defensive strength. He used the results of every National Football League game from the 1993 through 2005 seasons (excluding the 2006 Super Bowl), a total of 3,316 games, as a starting point.
For one thing, the data show a positive correlation between a team having a good offense and a team having a good defense. This makes sense. “A team with a good defense will cause more turnovers, will actually score more themselves, and create better field position,” Berry says. “A good offense forces opponents to play riskier strategies, create better field position, and can leave their defense on the field for shorter amounts of time.”
The following table lists the 10 best offenses and defenses of the last 13 years, according to Berry’s model. Each team has an offensive parameter (O) and a defensive parameter (D). The mean number of points scored by one team, S, is their offensive parameter, O, times the defensive parameter, D, for their opponent. The ratings are scaled so that a team with an average offense and defense playing another average team would each score 21 points.
Best Offenses |
Best Defenses |
|||||||
| Team | Year |
O |
D |
Team | Year |
D |
O |
S |
| Vikings | 1998 |
34.24 |
0.91 |
Ravens | 2000 |
0.621 |
22.42 |
21.26 |
| Chiefs | 2004 |
32.93 |
1.13 |
Buccaneers | 2002 |
0.622 |
25.72 |
20.48 |
| Colts | 2004 |
32.69 |
1.01 |
Chiefs | 1997 |
0.684 |
23.44 |
22.36 |
| Colts | 2003 |
32.27 |
0.99 |
Titans | 2000 |
0.686 |
21.49 |
22.14 |
| Rams | 2000 |
30.93 |
1.45 |
Patriots | 2004 |
0.688 |
26.98 |
21.28 |
| Broncos | 1998 |
30.80 |
0.87 |
Packers | 1996 |
0.703 |
27.81 |
21.65 |
| 49ers | 1993 |
30.44 |
0.96 |
Chiefs | 1995 |
0.708 |
21.85 |
21.55 |
| 49ers | 1994 |
30.00 |
1.04 |
Bills | 1999 |
0.714 |
21.11 |
21.42 |
| Chiefs | 2002 |
29.54 |
1.00 |
Colts | 2004 |
0.714 |
19.31 |
21.09 |
| Rams | 2001 |
29.43 |
0.92 |
Browns | 1994 |
0.715 |
20.69 |
21.04 |
As evident from the table, Berry’s model suggests that the 1998 Minnesota Vikings had the best offense of the last 13 years, with O = 34.24. The 2000 Baltimore Ravens had the best defense, with D = 0.621.
To get an estimate of the likely outcome of a game between two teams, multiply one team’s offensive parameter by the other team’s defensive parameter. In Berry’s model, an average team would be expected to score an average of 21 points. So, if the 1998 Vikings were to play the 2000 Ravens, the Vikings would be estimated to have a slight edge, scoring 21.26 points (O times D).
Overall, the best offense appears to be a slightly better bet than the best defense. Only one pairing in the top 10 gives the defense a slight advantage (the 2004 Kansas City Chiefs versus the 2002 Tampa Bay Buccaneers).
The trend also holds for superior and average teams. “Though certainly not convincing, there is evidence that a good offense beats a good defense (all else being equal),” Berry concludes. “This differs from the common perception that defense wins championships.”
But for really awful teams, he says, “it does appear that a bad defense beats a bad offense—in fact, a really bad defense is a big advantage over a really bad offense.”
Of course, offense and defense are fundamentally inseparable. What happens on offense affects what happens on defense and vice versa.
Moreover, statistical models, no matter how sophisticated, can’t say with certainty which team will win any given game. There’s enough variability in the game of football to make all sorts of outcomes possible—even ones that appear to be highly unlikely.
Indeed, it’s interesting to note that, of the top offensive teams of the last 13 years, only two actually went on to win a Super Bowl (1994 Forty-Niners and 1998 Broncos). Of the top defensive teams, four won the Super Bowl (2000 Ravens, 2002 Buccaneers, 2004 Patriots, and 1996 Packers). And two of the top offensive teams (2002 and 2004 Chiefs) didn’t even make the playoffs.
As for the 2005 Super Bowl, Berry’s model gave the Seattle Seahawks (better defense and slightly worse offense) the edge over the Pittsburgh Steelers. Pittsburgh won.
