Measuring impact in basketball | Plus minus & RAPM

Have we found the Holy Grail of basketball stats? In Part 5 of the Thinking Basketball stats series, we dive into the entire world of plus-minus as a way to determine impact or value. Is single-game plus-minus any good? Full season vs. on/off data? How do we account for teammates? And if we make the right adjustments, do we have an ideal one-number stat for ranking NBA players?

Stats series:

Additional resources:

Rob Antle summarizing RAPM:

Daniel Myers on APMs:

Joe Sill’s paper using Ridge Regression:

Justin Jacobs series on Adjusted Plus-Minus (RAPM)
Deep Dive on Regularized Adjusted Plus-Minus I: Introductory Example
Regularized Adjusted Plus-Minus Part III: What Had Really Happened Was…
Exercising Error: Quantifying Statistical Tests Under RAPM (Part IV)

WOWYR primer:

Support at Patreon:
Podcast: (available on iTunes, Spotify, etc.)
Twitter: @elgee35

Ben Taylor is the author of Thinking Basketball, a Nylon Calculus contributor, creator of the Backpicks Top 40 series & host of the Thinking Basketball podcast.

Stats courtesy: @bballport

Player art by @CrumpledJumper

Music: Music by Jordan Ciurcovich (Forest) & The Way Forward, 80s Hairlight (Instrumental)


Add Yours →

As a Canadian who knew next to nothing about basketball and started to tune in when the Raptors made their run, your videos have been essential in helping me understand this game. Thanks so much for these videos

Do an episode about the Hollinger gamescore statistic….jimmy highroller did a bit recently about it and Jordan and hardens high all time rankings for the star and speculating how wilt would have all the top spots for the star had blocks and other stats been tracked back then….but it is an all encompassing stat similar to PER or box plus minus. Please check it out and do what you can. I'm eager to learn more and get your take on the metric as I had never heard of it before that video but it's apparently been around for awhile

don't think a player's value will ever be captured by a single stat. you might get a good idea of how someone played in a single game through a stat sheet, but film needs to be utilized too. if you never watched a game and just saw stats from warriors games, you'd never know how easy steph curry made the game for KD.

I would love to some sort of advanced "gravity" statistic, measured by like help steps per drive or something like that. I'm not exactly sure how that would work but it would be interesting to see how many people require attention away from their primary assignment

This was a fascinating discovery of a classic idea from Boxing and MMA called "boxing/mma math" the idea of judging a fighter based of a long line of defeats and wins to say one fighter is better than another, X beat Y beat F beat G beat A, therefore X is better than A.

This is general viewed as very poor thinking and usually shown to be totally wrong in practice when the result of fights that can be put in this kind of chain having basically the same win loss ratio for both sides. Despite the appearance of a logical chain of thought process. Which makes me think that even if you could go down the line of Jordan-Meyers-etc. It wouldn't actually give you an accurate idea of skill or impact.

This is exactly why I keep saying the goat convo is pointless. All the greats have done things that only they could have done. Could Lebron have won 6 rings in the 90s? Probably not. Could Jordan have taken trash teams to the finals year after year? Probably not

Leave a Reply