Nylon Calculus: Using NBA matchup data to define defensive roles
By Jared Dubin
This post is a collaboration with Krishna Narsu.
The statistical revolution that has swept through basketball over the last decade-plus has primarily served to provide people in and around the NBA better tools with which to quantify players’ contributions on the offensive side of the floor.
There have been some incremental advances in measuring players’ defensive abilities, to be sure; but for the most part, the all-in-one defensive metrics that have taken hold over the years are relatively limited in their ability to describe how a player impacts the game defensively. In other words: while most of these statistics are able to give fans, coaches, and executives a decent idea of which players might be “good” defenders, they lack appropriate context and they almost never answer why those players should be considered good at defense. They don’t really describe what specific defenders actually do — let alone what they do well.
Is a player great at help defense? Does he pressure the ball and drain seconds off the shot clock? Is he a guy you can sicc on the opposing team’s best player in order to make that guy work as hard as possible for his points? Statistics like Defensive Win Shares, Defensive Box Plus-Minus, Defensive Real Plus-Minus, and the like don’t really answer those kinds of questions. Meanwhile, smaller-scale metrics capable of showing how good particular players are at specific acts (such as defending shots at the rim or deflecting passes) don’t always provide enough context to explain what kind of effect those skills have on team performance.
There are all different kinds of good defenders, but unfortunately, it’s always been difficult to tease out (and show supporting evidence of) which kind of defender a player is, given the limited data we’ve had available. Recently, however, the NBA released a trove of “matchup data” that had previously not been available to the public. Teams have been using it for awhile and have likely gained some valuable insights from it, but for the most part, the public had only gotten snippets of this data that had been privately provided to larger outlets and authors of Sloan Sports Analytics Conference research papers. Now that everyone has access, we can finally dig in and learn a little bit more.
Read More: Playoff implications of the full-strength standings
And while we are still likely a ways away from definitively stating which players are the best defenders and why, we can take small steps in that direction. By analyzing the matchup data, we can shine a light on exactly how defenders are being used by their coaches — and have more than just what we see with our eyes to support those assertions.
Consider the way an NBA offense works. Most teams have one player that serves as an initiator of the action, another player or two that carry the scoring burden, and two or three supplementary players that largely act as screeners, spot-up shooters, or tertiary off-dribble scoring threats.
Analyzing the new matchup data in conjunction with other publicly available metrics allows us to discern which players spend the most time defending initiators and/or primary scoring options. Using a variety of metrics (Time of Possession Percentage and/or Touches per Minute for initiators; Usage Rate and/or Offensive RPM for primary options), we can rather easily identify which players occupy the Initiator and Primary Option roles for each NBA team. We can then utilize the matchup data to tell us which defenders guard those types of players most often, as a percentage of their defensive possessions.
One interesting place an analysis of this data set takes us to is Boston. The Celtics had the NBA’s best defense for much of the season, a fairly incredible feat considering how much time was missed by some of their best defenders. Gordon Hayward, an excellent wing defender, went down a few minutes into the season. Marcus Smart played only 54 games. Jaylen Brown and Daniel Theis missed double-digit games as well. Much of the credit for the Celtics’ defense has rightfully gone to Al Horford — a fringe Defensive Player of the Year contender — and the team’s overall system. Brad Stevens did an excellent job coaxing better-than-expected defensive seasons from nearly every Celtics rotation player. A relatively unsung member of the defensive core, however, is backup guard Shane Larkin.
He was easily the Celtics’ most under-the-radar acquisition last summer, but the way Brad Stevens used him during the year made that acquisition count. Boston had a 99.0 defensive rating with Larkin on the floor this season, per NBA.com, the second-best mark on the team (among rotation regulars) behind only Aron Baynes. Larkin’s heavy ball pressure on opposing Initiators played a huge role in that success.
Surprisingly, it was Larkin — not Smart or Kyrie Irving or even Terry Rozier — that defended the opposing team’s leader in both Time of Possession Percentage (TOP%; the percentage of time a player has the ball in his hands while in the floor, which can be gleaned from Second Spectrum’s time of possession data) and Touches Per Minute for the greatest percentage of his defensive possessions. Larkin defended those players on more than 25 percent of his possessions, a rate on par with players like Dejounte Murray and Russell Westbrook (i.e. starting point guards), despite the fact that Larkin started only two games all season and played more than 30 percent of his minutes in garbage time.
This is largely because Stevens almost always reserved his close-game usage of Larkin for times when he needed the diminutive guard to give all-out effort defensively in order to delay an opponent’s sets and force them into turnovers. As a result, opposing teams had a 16.7 percent turnover rate against the Celtics with Larkin on the floor this season, a full 1.6 percentage points higher than the team’s next-closest regular rotation player (Smart).
We’ve already been seeing similar results during the playoffs. With Irving out for the playoffs and Smart still recovering from UCL surgery, more of Larkin’s relentless pressure has paid dividends for Boston as he has helped Rozier harass Eric Bledsoe through the first two games of the series. Larkin has played 37 minutes across those two games, and the Bucks have a putrid 100.5 offensive rating with him on the floor, having turned the ball over on a truly embarrassing 18.3 percent of their possessions.
But Larkin, for all his merits, is still a something of a bit player, even while he’s acting as Boston’s full-time backup point guard due to injuries. And what most people are interested in is the stars. Luckily for us, there are an awful lot of interesting nuggets available when it comes to which players are defending the opposing team’s stars, as well.
An unsurprising group including Avery Bradley, Michael Kidd-Gilchrist, and Andre Roberson sits at or near the top of the list of players that spent the most time defending Primary Options (the opposing team’s leader in Usage Rate and/or Offensive RPM) during the regular season, but more notable than those types of defensive role players was the position on the list of Jimmy Butler—especially when compared to other stars like himself. Butler defended the opposing team’s Offensive RPM leader on 21.1 percent of his defensive possessions this year, good for 11th in the NBA. He was the only All-Star to rank in the top 15, and one of only three inside the top 50. Klay Thompson at 19.3 percent (27th) and Paul George at 17.7 percent (39th) are the others.
Butler has continued to work against the opposition’s top offensive option during the playoffs, drawing most of the early work against the Rockets’ James Harden. You might think that George is taking some time off by defending Joe Ingles rather than Donovan Mitchell, for the most part; but even though Mitchell led the Jazz in usage rate during the regular season, it was actually Ingles who led the team in Offensive RPM. So, George’s primary matchup during the early part of OKC’s series still has him doing important work.
Comparing the regular season defensive usage of that trio to other swingman All-Stars like Bradley Beal (13.7 percent), James Harden (12.9 percent), Victor Oladipo (11.8 percent), Kevin Durant (11.7 percent), DeMar DeRozan (10.6 percent), and LeBron James (10.0 percent) is an interesting exercise, especially as it confirms just how rarely James defended the opposing team’s top option this season. That number has mostly held steady for LeBron during the playoffs. He guarded Oladipo on 9.6 percent of his defensive possessions in Game 1 and 10.4 percent in Game 2. (Oladipo has barely spent any time covering James, largely because, well, he’s just not big enough.)
But while Butler defended those players on a greater percentage of his defensive plays than any other star, no player in the league guarded the opposing team’s Primary Option — whether determined by usage rate or Offensive RPM — for more total possessions during the regular season than the Pelicans’ Jrue Holiday. Holiday appeared in 81 of the Pelicans’ 82 regular season games, averaging 36.1 minutes a night. Only one NBA player (LeBron) was on the floor for more defensive possessions. But even though LeBron played over 200 more possessions than Holiday, Jrue defended the opposing team’s Primary Option more than twice as often. And not only did Holiday defend those players twice as often as LeBron; he did so on almost 200 possessions more than any other player in basketball.
All that experience defending the opposing team’s best scorer has paid off for Holiday and the Pelicans during the playoffs. Jrue has spent most of the Pelicans’ two early-series wins tangling with Damian Lillard, and it’s difficult to imagine someone doing a better job on an opposing star than Holiday has thus far. According to the matchup data, Holiday has defended Lillard (Portland’s regular-season leader in both usage rate and Offensive RPM) on 47.4 percent of his defensive possessions—a rate that would have led the league by a mile.
Next: How often do we see 'Vintage Dwyane Wade'?
The Blazers as a team have scored just 82.4 points per 100 possessions with Holiday defending Lillard, a rate that would rank dead last in the history of the league. Dame himself has just seven points on 2-of-18 shooting with Holiday as his primary defender on the possession, compared to a significantly better 28 points on 11-of-23 shooting against other defenders. Portland has also scored a scorching 114.0 points per 100 possessions with non-Holiday players defending Lillard for the majority of the possession, so you can plainly see the impact Jrue’s defense has had.
Of course, we all know that on-ball defense is not the only thing that matters. Players do a ton of work to guard the opponent away from the ball; and that work is often just as critical to preventing the ball from going into the basket. In the next post in this series, we’ll dig into the players doing the most work defending opponets that use screens away from the ball to get themselves open.