Skip to content

Calculating Regularized Adjusted Plus-Minus (RAPM) with R

Notifications You must be signed in to change notification settings

tonyelhabr/nba-rapm

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

79 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Introduction

This primary goal of this project is to calculate Regularized Adjusted Plus-Minus (RAPM)—an “advanced statistic”—for NBA players. The calculated values can be found in the set of metrics_join CSVs in the project’s repository.

I plan to write about this project in more detail on my blog. so I encourage the reader to read more about it there.

Usage

If you were to fork this project and try to run it from scratch, below shows the required order of function calls.

First, download all of the data needed.

# pre-process ----
# Note that `overwrite = FALSE` is the default, but setting it explciitly here to remind
# the user that it is an option.
# This goes to the
download_pbp_raw_files(overwrite = FALSE)

download_nbastatr(overwrite = FALSE)
download_rpm_espn(overwrite = FALSE)
download_rapm_sz(overwrite = FALSE)

Next, run the “main” function. This is what is run with the command-line interface (CLI) that also comes with the project.

# This reads from the config.yml files.
auto_main()

Highlights

Below is a visual comparison of various RAPM-related metrics, either calculated in this project (i.e. calc) or retrieved from an external source.

The data behind this visual

y apm_calc bpm_nbastatr dbpm_nbastatr drapm_calc drapm_sz drpm_espn obpm_nbastatr orapm_calc orapm_sz orpm_espn pm_nbastatr rapm_both_calc rapm_calc rapm_sz rpm_espn
apm_calc NA 0.0857897 0.0068202 0.3504192 0.1118109 0.0059843 0.0863602 0.4191907 0.0972412 0.0633819 0.1194202 0.5170582 0.6750558 0.2135919 0.0727325
bpm_nbastatr 0.0857897 NA 0.2852047 0.0273031 0.0027746 0.0471514 0.8274349 0.1482843 0.0453781 0.1901469 0.0437862 0.1743086 0.1600988 0.0418785 0.2560962
dbpm_nbastatr 0.0068202 0.2852047 NA 0.0912274 0.0503027 0.2727653 0.0168123 -0.0024059 -0.0017948 -0.0018229 0.0191676 0.0321596 0.0191713 0.0189409 0.1234330
drapm_calc 0.3504192 0.0273031 0.0912274 NA 0.3322217 0.0765407 -0.0020656 0.0098078 -0.0003777 -0.0027973 0.1413987 0.4035084 0.3356299 0.1240432 0.0358574
drapm_sz 0.1118109 0.0027746 0.0503027 0.3322217 NA 0.1041948 -0.0009822 -0.0023232 -0.0017017 -0.0006295 0.1791320 0.1442801 0.0756311 0.4384602 0.0256227
drpm_espn 0.0059843 0.0471514 0.2727653 0.0765407 0.1041948 NA -0.0008774 -0.0029342 -0.0026649 0.0064440 0.0491640 0.0298905 0.0175993 0.0421981 0.3028806
obpm_nbastatr 0.0863602 0.8274349 0.0168123 -0.0020656 -0.0009822 -0.0008774 NA 0.2368059 0.0650393 0.2843836 0.0300633 0.1443483 0.1512988 0.0278430 0.1757929
orapm_calc 0.4191907 0.1482843 -0.0024059 0.0098078 -0.0023232 -0.0029342 0.2368059 NA 0.3812504 0.1694076 0.2074359 0.4420484 0.7619807 0.2037157 0.1178648
orapm_sz 0.0972412 0.0453781 -0.0017948 -0.0003777 -0.0017017 -0.0026649 0.0650393 0.3812504 NA 0.1361238 0.3027395 0.1785147 0.2277643 0.5434344 0.0906094
orpm_espn 0.0633819 0.1901469 -0.0018229 -0.0027973 -0.0006295 0.0064440 0.2843836 0.1694076 0.1361238 NA 0.1209535 0.1041428 0.1121011 0.0596036 0.6037412
pm_nbastatr 0.1194202 0.0437862 0.0191676 0.1413987 0.1791320 0.0491640 0.0300633 0.2074359 0.3027395 0.1209535 NA 0.3326310 0.3124833 0.4888706 0.1893420
rapm_both_calc 0.5170582 0.1743086 0.0321596 0.4035084 0.1442801 0.0298905 0.1443483 0.4420484 0.1785147 0.1041428 0.3326310 NA 0.7345316 0.3308535 0.1489425
rapm_calc 0.6750558 0.1600988 0.0191713 0.3356299 0.0756311 0.0175993 0.1512988 0.7619807 0.2277643 0.1121011 0.3124833 0.7345316 NA 0.2980069 0.1374234
rapm_sz 0.2135919 0.0418785 0.0189409 0.1240432 0.4384602 0.0421981 0.0278430 0.2037157 0.5434344 0.0596036 0.4888706 0.3308535 0.2980069 NA 0.1136859
rpm_espn 0.0727325 0.2560962 0.1234330 0.0358574 0.0256227 0.3028806 0.1757929 0.1178648 0.0906094 0.6037412 0.1893420 0.1489425 0.1374234 0.1136859 NA

Top 20 RAPM players for 2017 (according to my calculations, which are probably off 😆)

name slug rank drapm orapm rapm
Dante Exum UTA 1 2.69 3.43 6.12
Stephen Curry GSW 2 -0.15 5.04 4.90
Marcus Georges-Hunt 3 -1.07 5.79 4.71
Jordan Bell GSW 4 1.00 3.39 4.38
Brandan Wright 5 2.66 0.99 3.66
OG Anunoby TOR 6 1.30 2.33 3.63
Iman Shumpert SAC 7 2.78 0.79 3.57
Chris Paul HOU 8 0.32 3.05 3.37
Nene HOU 9 1.04 2.16 3.20
Mike Conley MEM 10 2.19 0.94 3.13
Eric Gordon HOU 11 1.04 2.07 3.11
Torrey Craig DEN 12 0.83 2.26 3.09
Lucas Nogueira 13 1.14 1.90 3.04
PJ Tucker HOU 14 0.76 2.22 2.99
Robert Covington MIN 15 1.09 1.88 2.97
Zaza Pachulia DET 16 0.31 2.57 2.88
Thabo Sefolosha UTA 17 1.02 1.83 2.85
Ersan Ilyasova MIL 18 1.02 1.57 2.59
Joel Embiid PHI 19 0.97 1.49 2.46
Ekpe Udoh UTA 20 1.00 1.46 2.46

2017 offensive RAPM coefficients for top 10 players as a function of cross-validated (CV) log-lambda values

2017 defensive RAPM coefficients for top 10 players

Ridge regression CV lambda penalties for 2017 offensive RAPM

Ridge regression CV lambda penalties for 2017 defensive RAPM

About

Calculating Regularized Adjusted Plus-Minus (RAPM) with R

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published