Dataset contains play-by-play data from three sources: stats.nba.com, data.nba.com and pbpstats.com and also shots details. Dataset contains data from season 1996/97 for stats.nba.com and shotdetails, from season 2000/01 for pbpstats.com and from season 2016/17 for data.nba.com.
Play-by-play data collected with help functions from advanced_pbp_data repository.
Detailed description data can be read in file description_fields.md.
Note: dataset contains only Regular Season data, without play-off
Ryan Davis - Analyze the Play by Play Data
Python nba_api package for work with NBA API - https://github.com/swar/nba_api
R hoopR package for work with NBA API - https://hoopr.sportsdataverse.org/
I made this dataset because I want to simplify and speed up work with play-by-play data so that researchers spend their time studying data, not collecting it. Due to the limits on requests on the NBA website, and also because you can get play-by-play of only one game per request, collecting this data is a very long process.
Using this dataset, you can reduce the time to get information about one season from a few hours to a couple of seconds and spend more time analyzing data or building models.
I also added play-by-play information from other sources: pbpstats.com (there is information about the time of ownership and the type of its beginning) and data.nba.com (there you can find coordinates of actions on court). This data will enrich information about the progress of each game and hopefully add opportunities to do interesting things.
You can download dataset several ways:
Clone git repository to your device
git clone https://github.com/shufinskiy/nba_data.git
Download using a programming language
You can write your own loading functions or use ones I wrote for R and Python languages.
R:
get_nba_data <- function(seasons = seq(1996, 2021), data = c("datanba", "nbastats", "pbpstats", "shotdetail"), untar = FALSE){
df <- expand.grid(data, seasons)
need_data <- paste(df$Var1, df$Var2, sep = "_")
temp <- tempfile()
download.file("https://raw.githubusercontent.com/shufinskiy/nba_data/main/list_data.txt", temp)
f <- readLines(temp)
unlink(temp)
v <- unlist(strsplit(f, "="))
name_v <- v[seq(1, length(v), 2)]
element_v <- v[seq(2, length(v), 2)]
need_name <- name_v[which(name_v %in% need_data)]
need_element <- element_v[which(name_v %in% need_data)]
for(i in seq_along(need_element)){
destfile <- paste0(need_name[i], ".tar.xz")
download.file(need_element[i], destfile = destfile)
if(untar){
untar(destfile, paste0(need_name[i], ".csv"))
unlink(destfile)
}
}
}Python:
import urllib.request
import tarfile
from pathlib import Path
from itertools import product
def get_nba_data(seasons=range(1996, 2022),
data=("datanba", "nbastats", "pbpstats", "shotdetail"),
untar=False):
if isinstance(seasons, int):
seasons = (seasons,)
need_data = tuple(["_".join([data, str(season)]) for (data, season) in product(data, seasons)])
with urllib.request.urlopen("https://raw.githubusercontent.com/shufinskiy/nba_data/main/list_data.txt") as f:
v = f.read().decode('utf-8').strip()
name_v = [string.split("=")[0] for string in v.split("\n")]
element_v = [string.split("=")[1] for string in v.split("\n")]
need_name = [name for name in name_v if name in need_data]
need_element = [element for (name, element) in zip(name_v, element_v) if name in need_data]
for i in range(len(need_name)):
t = urllib.request.urlopen(need_element[i])
with open("".join([need_name[i], ".tar.xz"]), 'wb') as f:
f.write(t.read())
if untar:
with tarfile.open("".join([need_name[i], ".tar.xz"])) as f:
f.extract("".join([need_name[i], ".csv"]),'./')
Path("".join([need_name[i], ".tar.xz"])).unlink()Download from Google Drive (without season 2022/23)
You can also download full version of the dataset from GoogleDrive.
If you have questions or proposal about dataset, you can write me convenient for you in a way.
