No tags yet.
NBA Five-Man Plus/Minus per Minute
Five-man plus/minus is a good stat to get a sense of which lineups are more/less productive. However, most of what is published is plus/minus per game. I think a better metric is five-man plus/minus per minute (FMPMpM). Looking at pus/minus on a per-minute basis evens the playing field for lineups that have been on the floor for minimal minutes. A lineup could be very efficient, but have a low per game plus/minus due to minimal minutes.
Being about a quarter of the way into the 2018/19 NBA season, many teams are still figuring out which lineups are/aren't working well together. I thought it would be interesting to take a deepish dive into FMPMpM, and see if any interesting early season patterns are emerging.
About the Data/Methodology
To calculate FMPMpM, I used play-by-play data from https://www.bigdataball.com. The data in this post covers every play for every game in the 2018/19 season through Friday, 11/21/2018. The BigDataBall dataset captures the 10 players involved in each possession of every game, the result, and the amount of time elapsed during each possession.
To calculate FMPMpM, we summarize the points scored for and against each lineup, along with the sum of the time for each possession. The final dataset looks like this:
A fair amount of work had to be done to clean and prepare the raw data for analysis. If you are interested in the process and R code associated with the data and visualizations in this post, you can find everything here. Following the narrative, code, and comments will provide you with a tutorial on several useful R techniques for data manipulation, summary, and visualizations with ggplot2.
Top 10 FMPMpM Lineups
Not surprisingly, the team with the best record (Toronto) has 2 lineups in the top 10. Interestingly, OKC with Dennis Schroder in the lineup is slightly better than with Russell Westbrook (Schroder is very under-rated). Seeing Anthony Davis in 2 of the top 10 is no surprise, but not seeing Giannis Antetokounmpo in the top 10 is, given his personal and team success so far this season. Personally, I am happy to see Sacramento doing well again. Buddy Hield and De'Aaron Fox have been really impressive so far.
Lineup Utilization by Team
Now let's look at a density plot to see the distribution of lineups for each team in terms of FMPMpM.
What these plots show is that most of the lineups are just over/under break even (0). Good teams generally have more density to the right of center (positive FMPMpM) than left. Examples are Toronto and Milwaukee. Bad teams will generally have more density to the left (negative FMPMpM). Examples are Atlanta and Cleveland. Teams with skinny peaks are mostly playing lineups that are not extremely good or bad.
Next, we'll dive into this a little deeper by plotting each lineup individually, comparing FMPMpM to minutes played.
Here we see that most lineups get very few minutes together. Only a few lineups for each team are getting big minutes. Minnesota looks to spread out their minutes more than most. By contrast, aside from 2 lineups, Atlanta is experimenting with several 5-man combinations. Orlando is giving higher minutes to more lineups than most.
The other thing we see here is the relationship between minutes and high/low FMPMpM. Most of the big minute lineups are close to the center. Keep in mind, this is a per minute metric, so + 0.1 FMPMpM vs. - 0.1 FMPMpM represents almost a 10 point difference when extrapolated to a full game. So, being barely on the right of 0 is significantly better than being barely to the left of 0.
The last thing we'll look at is the same plot as above, but limited to only lineups that have played at least 48 minutes together. Forty-eight minutes equates to a full game. This will show how the preferred lineups of each team is performing.
As we saw in the prior graph, you can see here that only a small number of lineups are given big minutes for each team. Orlando spreads the wealth more than any other team with 4 lineups getting 48+, and three of their four are making a very positive impact. I hate to keep picking on Atlanta, but their 2 preferred lineups are terrible. This is what a graphical representation of tanking looks like. Toronto, OKC, New Orleans, Philly, Charlotte, Memphis, Milwaukee, and Sacramento are all doing well to limit their big minute lineups to those making a positive contribution.
Wrapping up
The season is just starting to take shape. As it continues, it will be interesting to keep an eye on FMPMpM as teams figure out which combinations work best together. Teams that are tanking will likely continue to stick with younger lineups that struggle, but the teams in contention should show a shift in minutes toward more productive lineups. We'll check back in a few weeks to see if/how things have changed.