In the Brad Stevens era, the Celtics have ranked in the top 5 for three-point defense each season. This is pretty incredible because three-point defense is notoriously volatile. It’s so unpredictable that many writers chalk it up to luck. The nerds at Nylon Calculus even said:
Teams have very little control over the opponent 3-point percentage. They can affect the number of attempts, but not the percentage.
Writers and pundits have shown skepticism specifically toward the Celtics’ three-point defense as well. Kevin Pelton’s model discounted Boston’s defense because of this expected regression. However, there is reason to believe that Boston’s three-point defense is driven by more than luck.
There are many factors that affect a three-point attempt, but this article looks specifically at where the shot occurs, how open the shooter is, and the quality of shooter.
Corner threes are the most efficient shot next to dunks. Teams shoot much better in the corner than they do from above the break, so allowing fewer of these shots should lead to a lower opponent 3P%. The Celtics ranked 4th in the NBA last year in the proportion of corner threes allowed. Boston’s opponents took 21.6% of their 3PA from the corners, compared to the league average of 23.8%. This isn’t a huge difference, but it certainly helps.
How open is the Shooter?
The NBA.com defense tracking data isn’t perfect, but it’s the closest thing we have to approximate open shots. I assumed there would be a huge difference in Boston’s wide open threes allowed compared to the NBA average. But, the Celtics allowed an almost identical proportion of shots across the board.
Distribution of Opponent 3PA
This is where drilling down into the data leads to some interesting results.
Who is taking the Shot?
Stevens has talked about the importance of shooter quality before.
I think we’re doing a pretty good job of identifying guys that make shots at a high rate... and we’ve been guarding the 3-point line good as a result
With that in mind, I sorted players into three categories based on their 2017 3P%.*
Shooter Quality Criteria
It’s not perfect, but should work as a rough way to “identify guys that make shots at a high rate”.
Filtering the data using shooter quality gives support to what Stevens was talking about. The Celtics allow fewer open and wide-open attempts from “good shooters” compared to the NBA average.
Looking at how opponents actually converted on these attempts is interesting too.
For the most part, players also shot worse compared to the NBA average. It’s a similar story with corner threes as well.
The Celtics limit the amount of corner threes taken by good shooters, and those players have a lower 3P% as well.
Balancing the Factors
Numbers show what happened, but they don’t answer every question. Why do good players shoot significantly worse on wide-open threes against the Celtics? Why do they shoot worse in the corners? A portion of that is definitely luck, but both scheme and defender ability should play a role too.
Brad Stevens has been talking about “defensive DNA” since his Butler days, stressing good close-outs, fighting through screens, and the utility of switching defenders when necessary. You see it in the way Marcus Smart defends and in the praise Jaylen Brown received from J.J. Redick. These little things may not directly affect 3P%, but they should bear out statistically if they’re consistently happening.
Most analysis on three-point defense focuses on volatility to say that teams can’t control it. That still seems broadly true. But after looking at Boston’s continued success, I believe teams can influence opponent 3P%. More specifically, I think the Celtics have been able to reduce the role that luck plays in opponent 3P%.
This is important because affecting something generally out of your control is a way to gain an edge over an opponent. In a league driven by stars, these marginal advantages are where coaching can shine through. So don’t be surprised if Brad Stevens and the Celtics continue their trend of great three-point defense this season.
*Big list of data concerns: no minimums set for 3PA, arbitrariness of “quality” cutoffs, season 3P% is imperfect measure of ability, other teams and prior years would help add context, etc.