Defense Wins Games

By Sean Carroll | November 14, 2006 6:23 pm

Bowen Blocks Dirk Matthew Yglesias and Tyler Cowen both consider the eternal question of whether defense or offense is more important, especially in the context of new NBA rules that allow for more scoring. Apparently some folks are arguing that, since it’s now easier to score, a team’s priority should be to bring in offensive-minded players, rather than concentrating on defense. (I’ll leave it as an exercise to the reader to identify the logical flaw there.) Yglesias argues that offense and defense must both be important, since the goal is to end the game with more points than the other team:

I concede that the new rules have made it harder to play defense. I fail to see, though, how that makes defense less important. Two factors determine who wins a basketball game: how many points your team scores and how many points the other team scores. Since you have the ball roughly half the time and the other team has the ball roughly half the time, it stands to reason that offense and defense should have exactly the same importance.

Unfortunately, that last bit is just as logically flawed as the previous argument. The truth is that defense is (still) significantly more important than offense in winning games.

How can that be, if teams (basically) spend the same amount of time, or number of possessions, on offense and defense? To decide which skill is more important, we have to consider the variation in results obtained by being good at one vs. being good at the other. In other words, which has a bigger effect on wins: being one of the best offensive teams, or being one of the best defensive teams?

Yglesias looks at some individual playoff results, which are somewhat inconclusive. But we can just look at the season stats and compare the results of being good at offense vs. being good at defense. Of course, we’re faced with deciding how to measure those skills. Points scored is actually not a good measure, since that is affected more by the pace of the game than by true offensive or defensive prowess. Points per possession would be perfect, but I don’t know where to find that stat. So instead let’s just look at Team Offensive/Defensive Field Goal Percentage (FG%), which is a pretty good proxy for offensive/defensive aptitude.

What you should really do is to type in all the data and correlate with wins, but that sounds like work. Instead, let’s just define a “good offensive/defensive team” as one in the top 10 of the 30 teams in the NBA in offensive or defensive FG%, respectively, and “bad” as being in the bottom 10. We immediately see that there is a greater range in defensive aptitude than in offensive aptitude. The median good offensive team shoots at a .474 clip, whereas the median bad offensive team shoots .439, for a difference of .035. But while the median good defensive team holds their opponents to .439, the median bad defensive team only holds their opponents to .478, for a difference of .039. In other words, there is a slightly bigger difference between good and bad defensive teams than good and bad offensive teams. Concentrating on defense, it should follow, would potentially have a bigger outcome in the win/loss columns.

And it does. The winning percentage of the good offensive teams is .580, while that of the bad offensive teams is .413, for a difference of .167. But the winning percentage of the good defensive teams is .615, while that of the bad defensive teams is .358, for a difference of .257. That’s a substantial difference. A good defensive team is much more likely to be a winner than a good offensive team.

The simple ex post facto explanation is just that all NBA players are pretty good scorers, or at least that the players who do the bulk of the scoring are all pretty good. There’s not too much of a difference in overall efficiency between the very-good and the truly excellent. But defensive abilities are much more variable, and perhaps also more dependent on coaching and team dynamics. Putting your effort into defense has a larger marginal payoff than putting it into offense. Which most coaches would agree with. People these days like to blame Pat Riley for that, but I think Bill Russell figured it out long ago.

  • Sourav


    Did you control for turnovers, poor free throw shooting and technical fouls? These may contribute to losses, and opponents’ FG % might simply correlate with them.

  • Sean

    Nope. But I encourage you to do so!

  • Elliot

    One note about the 6 time world champion Chicago Bulls. Michael Jordan and Scottie Pippen along with their offensive skills were both superb defensive players.

  • Jeff

    yeah, yeah…but defense is boring!

    j/k. The actual problem is that bad offense is boring. Good defense can be thrilling.

  • Nate

    You are making the assumption that creating a better than average defensive team costs the same amount as creating a better than average offensive team. That is not necessarily true. You give the payout in winning % increase is 1.5 to 1 (assuming no sample size problems), but maybe the cost ($, time, coaches, less exciting team with reduction in marketability) of a good defense is more than 1.5 to 1.

    Also, if you are using 06/07 data, that comes to 7 whole data points for each team! Maybe some sample size issues there? Looking at 05/06 in full would be better.

  • Joe Grossberg

    IMHO, the real answer is one level deeper, and something you touch on.

    Good offense may or may not result from good teamwork. But good teamwork is a prerequisite for good defense.

    Also, I think the PPG stats are too simplistic. Let’s say team A wins 120-100 and then loses 110-100; team B beats opponents by 104-100 and 106-102. By that measure, team A is superior in points for (115 vs. 105) and points allowed (100 vs. 101), yet has a worse record.

  • Dan

    You can get offensive and defensive points per possession data here. I didn’t bother to run the correlations with wins (since points are so closely related to wins), but last time that Matt made this argument I found that offensive and defensive PPP have approximately equal standard deviations over the last two seasons, 3.05 and 3.55 for offense vs. 3.49 and 3.27 for defense.

  • Urbano

    Sean, it is related somehow also to NBA, since onminipresence is something I expect of it :-): will you say nothing about that??

  • Eric

    The well-known saying in fact says it all:
    “Offense wins (regular-season) games; defense wins championships.”

    Your argument based on variation of results from offense vs. variation of results from defense is legitimate regarding regular season performance and is enhanced in the playoffs, since all playoff teams have at least a couple good offensive players. But the real root of the effect is that playoff games often come down to the final play(s) because most of the time playoff teams are fairly evenly matched.

    Now a team’s shooting performance on a given night (or in a given play) is largely a matter of chance, but playing good defense is closely tied to effort. Thus strong defensive teams can directly control their (defensive) performance, and in particular they can be sure that they perform their best when it matters most–in the playoffs and at the end of games. This is not true of offensive teams, who are at the whim of Lady Luck.

    In stupid sportscaster talk, championship teams are thus those “who can get stops when they need them.” Unfortunately I learned this first-hand a few years ago as a fan of the Sacramento Kings, who perenially lost nail-biters to the Los Angeles Lakers because of L.A.’s superior defense.

    (By the way, note that there is the coupled issue of whether it’s better to have a couple superstars or a bunch of good players, but that merits its own discussion.)

  • Nate

    Well, I decided to do all the “heavy lifting” for you guys to run the correlations of FG% vs Win % offense and defense. (It involved not only 2 copy-paste clicks, but a selection of “add trendline” in the excel graph for the OLS — whew!)

    For the whole 05/06 season: 1 % higher fg on offense = 4.5% higher win. 1% lower fg on defense = 4.6% higher win.

    Looks the same to me. One interesting note though, the R-squared is better for defense (.17 off vs .31 def). (which doesn’t mean the effect is greater, just that is may be a little more predictable from one team to the next).

    In the end, I still think the issue is that teams don’t just create fg%, they must pay players to create them. So the costs of being more offensive-minded (getting iverson) or defensive-minded (getting wallace) are the primary driver. How to get the most bang for your buck?

  • r stevens

    Consider the fact that a better defensive team not only
    lowers the opponents f.g.%,but also creates more defensive turnovers;thus more offensive scoring opportunities,for their team;less for the opposition.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Cosmic Variance

Random samplings from a universe of ideas.

About Sean Carroll

Sean Carroll is a Senior Research Associate in the Department of Physics at the California Institute of Technology. His research interests include theoretical aspects of cosmology, field theory, and gravitation. His most recent book is The Particle at the End of the Universe, about the Large Hadron Collider and the search for the Higgs boson. Here are some of his favorite blog posts, home page, and email: carroll [at] .


See More

Collapse bottom bar