I am obsessive about Basketball. It’s basically my religion.
Every week, on Saturday, I head to Xavier High School on 16th street for my equivalent of church – a weekly pickup run with guys 10 years older than me. It is from the elders that the truths of the game are passed down: give the ball up to get it back, set a screen to free yourself up, team basketball wins on both ends.
The truths from my church are largely the same truths at basketball’s equivalent of a megachurch: The NBA. At the highest level, the “Meta” of the game has changed immensely since the mid-2000s. While historically, basketball has been a game dominated by scoring from as close to the basket as possible, the invention of the three-point line in the 1980s and its popularization by Mike D’Antoni and the “7 Seconds or Less Suns” has radically changed the way the game has been played over the past 15 years.
In this project, I am to look at the current state of NBA lineups and understand their compositions. I would like to visually represent the current “meta” of the NBA in terms of lineup composition – the teams that front offices are constructing and coaches are utilizing on the floor. Then I would like to look at commonalities in lineups that are effective vs. ineffective (measured with Net Rating).
My audience with this project is basketball nerds. The type of dorks on reddit who 8 years ago were arguing if Nikola Jokic would be an MVP level player. The same ones arguing about Isaiah Hartenstein right now. The ones trying to figure out which “advanced” statistics are the best for measuring defensive impact. The ones who will argue that the eye test is the only thing that matters on that side of the floor. Basketball, like any other subculture, has a vernacular – much of this project will be operating within that vernacular. As simply as possible, I will try to decode what this terminology means.
To address my question(s) about the current “meta” of the NBA, I will be looking at 5-man lineup combinations that have played for a minimum of 100 minutes together. There are 104 lineups across 30 teams that have hit this mark. From there, I will classify each player in these lineups into one of three positions: Guard, Wing, and Big. Traditionally, there have been five positions used – but in the modern meta, there are three main roles that are filled. From there, I will look at the number of 35% three-point shooters in each lineup (league average) and number of below-average defenders in each lineup (measured in DEPM).
The first tab of the story discusses the current “Meta” of the league at large. In the first graph, there is a bar chart that reflects the number of lineups each team used that played at least 100 minutes. The color scheme associated with each team reflects the number of wins the team had this year.
The graph in the top right quadrant of the story reflects the different types of lineup compositions in the NBA, split into the three roles that players tend to fall into: guards, wings, and bigs. These days, the most common lineup played is 2 guards, 2 wings, and 1 big, followed closely by 2 guards, 1 wing, and 2 bigs. 10-15 years ago, there were hardly any lineups without two bigs. While the “spacing” revolution of the NBA has been widely discussed, it is interesting to see that reflected clearly in lineup compositions. This chart is colored by net rating, and interestingly enough, the most effective combo across the league was 1 guard, 3 wings, and 1 big – though it was only found in 7 lineups.
The bottom graph of the first story breaks out every lineup that each team played for a minimum of 100 minutes and reflects the net rating in the shading for each lineup. It’s interesting to see how many lineups that were net negatives played a significant sample size over the course of the season. My favorite lineup to look at is on the Golden State Warriors: a three guard lineup of Steph Curry, Klay Thompson, & Brandin Podziemski, flanked by Jonathan Kuminga and Kevon Looney. Two hall-of-famers, yet they were one of the worst lineups in the league.
The second tab of the story looks at the number of 3-point shooting threats in each lineup. “Shooters” (35%+ from 3P) and “Great Shooters” (40%+ from 3P) both have extreme impacts on the offensive effectiveness of lineups. Unsurprisingly, lineups without multiple shooting threats struggle to score efficiently as defenses are able to “pack the paint” – to leave non-shooters alone away from the basket, and double-team the player with the ball. This is reflected in both of the top graphs of the second story, which displays Offensive Efficiency (measured in Offensive Rating) in the color.
It’s interesting as well to see that not a single lineup has more than 3 players who shoot above 40% from 3… I wonder if that is to come soon.
Also – looking at the “Shooters per Lineup & Net Rating” chart, I thought that there might be some decline in the defensive effectiveness of a lineup that could be seen in the Net Rating. No such trend was found. The more shooters on the floor, the better.
The final tab of the story looks at the number of below-average defensive players found in each lineup (measured in DEPM). As shown in the graphs, the more poor defenders found in each lineup, the worse both the Defensive Rating & the Net Rating are. Interestingly enough, the best Net Rating lineups on average have one below-average defensive player. I believe a conclusion that one can draw here is that a good defensive team can hide one weak point (that is super effective on offense). Two weak points? Diminishing returns.
Throughout the project, I used tree maps because I find them very effective for displaying data with a few categories & multiple dimensions. When discussing either the number of shooters or poor defenders in a lineup, it splits the data quite nicely into sized blocks that reflect color more cleanly than bar graphs.
In the future, I would like to go back to the NBA archives and look at past seasons. When I just checked, the oldest data is from 2007-2008, so it would be fascinating to see how the league’s meta has changed since then!
