Over the summer, I embarked on an exercise to determine how much control offense and defense had on various stats in a particular game. There were 14 posts in all, plus a bonus post on how one could misuse stats to show that the defense has control over free throw percentage. If you read all of those posts, you have my fullest admiration, but you are also an unusual person. For the rest of you, herein lies a summary of those works.

For starters, my purpose here was to get a better understanding of what the box score means. On Saturday, Wright State made 5-of-32 two-point shots against George Mason. I didn’t see the game. I know nothing about what happened. How much of that wretched performance can be blamed on Wright State’s offense and how much can be credited to George Mason’s defense? And even if I did see the game, having a background knowledge of what the offense or defense can typically control can be useful.

To do this, I made a very simple model to predict a stat for a game based on the offense’s and defense’s ability with respect to said stat. For instance, how well does a team’s offensive free throw percentage and its opponent’s defensive free throw shooting predict a team’s free throw shooting in a game? Naturally, the defensive value isn’t useful for predicting free throw percentage, but for other stats it gets more interesting. You can read more about the methodology here. And links to the other parts of this series are listed at the end of this piece.

Here’s the master list of stats and the percentage of influence the offense has over them in a particular game along with the home court advantage associated with each stat. (APL= average possession length. NST% = non-steal turnover percentage. I hope the others are familiar to you.)

FT%  98%  (HCA=+0.5%)
APL  86%  (HCA=-0.1s)
3P%  83%  (HCA=+0.7%)
OR%  73%  (HCA=+1.1%)
3PA% 71%  (HCA=0.0%)
A%   71%  (HCA=+2.6%
PPP  64%  (HCA=+3.7)
NST% 59%  (HCA=-0.4%)
2P%  50%  (HCA=+1.4%)
TO%  49%  (HCA=-0.7%)
FTR  36%  (HCA=+2.8)
Stl% 30%  (HCA=+0.4%)
Blk% 15%  (HCA=+1.2%)

Here are a few of my favorite takeaways:

A good offense beats a good defense. A bad offense will struggle against against a bad defense. In the college game anyway, offense has majority control of scoring. That is to say, if we know the quality of a team’s offense and its opponents defense, we should expect the offense’s efficiency in a game to be closer to the offense’s ability than the defense’s.

The offensive control over shooting percentage decreases the closer you get to the basket. We’ve been through the issue of 3-point shooting many times, but this approach again confirmed that the defense’s ability to affect opposing three-point percentage is quite limited. However, two-point percentage is roughly an equal battle between offense and defense. In other words, our initial assumption should be that Wright State’s dismal performance was equal parts offensive incompetence and defensive excellence. While we are only looking at 2’s and 3’s, one can imagine that long 2’s are more in control of the offense since those shots share some similarities with 3’s. Basically, an offensive player is more likely to take a shot in the presence of a defender the closer he is to the rim.

Two-point percentage and turnover percentage are the battlefield stats. Two-point percentage and turnover percentage are under nearly equal influence of the offense and defense. Other stats are in control (to varying degrees) of the side providing the action for the stat. Blocks, steals, and fouls are under defensive control, while everything else is under offensive control.

There’s a ton of variability in shooting percentage. I embarked on this adventure just looking at the percentage of control the offense had compared to the defense. But there are really three pieces to the predictive pie…offense, defense, and random variance. So while the offense is basically in total control of its own free throw percentage, on a game level there’s a bunch of variance. Even the best free throw shooting teams are going to have off nights. And to a slightly lesser extent, this idea applies to three-point shooting. Two-point shooting is quite a bit more predictable.

And that leads me to a different way of looking at the control issue. Below are the stats ranked by each side’s absolute control over them.

Offense  Defense
APL      Blk%
3PA%     PPP
PPP      3PA%
A%       Stl%
TO%      TO%
2P%      FTR
NST%     2P%
OR%      APL
Stl%     A%
FTR      NST%
FT%      OR%
Blk%     3P%
3P%      FT%

So while the defense only has 29% control of its opponent’s three-point attempt percentage, in terms of absolute control there are only two things it has more infleunce over – block percentage and points per possession. There is less random variance associated with three-point attempt percentage than any other stat I looked at except for points per possession. And the only thing the defense has less absolute control over than opposing three-point percentage is opposing free-throw percentage.

Three-point shooting is the great equalizer for the offense. Three-point percentage isn’t influenced much by the defense and it is also more immune to the location of the game than two-point percentage. Home court advantage for three-point percentage is 0.7% whereas for two-pointers it’s 1.4%. I don’t have any evidence that teams dependent on three-point shooting and one has to remember that there’s a lot of variance in it from game to game. But over the long run, three-point shooting is more likely to hold up against opposing defenses and opposing crowds than two-point shooting, so it’s not a crazy theory.

The defense’s tools are two-point defense and influencing shot selection. While a frightening number of things are in the offense’s control – leading to the offense having 64% control over its points per possession number – the defense has significant influence over where shots are taken from and how effective the offense is near the rim.

These numbers are averages across all teams. It’s important to keep in mind that the numbers presented here don’t apply equally to all teams. But I would use these numbers as a default. A team that decided to never contest a three-point shot is going to see its opposing 3P% number go up in the long run. It’s possible that an offense with Blake Griffin or Adam Morrison is more in control of its foul rate than most teams.

I know it’s difficult, but there are no absolutes to be taken from this work. In fact, I don’t know how I’d use this information strategically. It was really meant to get a handle on how to interpret a random box score. You might think a defense should make it a priority to limit opposing three-point attempts, and in certain cases that would be advisable. But this approach could have implications for a team’s two-point defense depending on its personnel (and how it goes about limiting three’s). So, like many things in basketball, it’s complicated.

Here’s the whole series:
Free throw percentage
Is free throw defense real?
Average possession length
Three-point percentage
Offensive rebounding percentage
Three-point attempt percentage
Assist rate
Points per possession
Non-steal turnovers
Two-point percentage
Turnover percentage
Free throw rate
Block percentage