In solmos/euroleaguer: Extract and Analyze Euroleague Basketball Generated Data

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

eurolig provides a toolkit to easily retrieve and analyze basketball generated data for the Euroleague. The package is mainly designed to work with two types of data: play-by-play data and shot location data. Although Euroleague's first season was in 2000, play-by-play and shot location data is only available since the 2007/2008 season. This vignette introduces the package (v0.4.0) and shows how it could be used to do some basic analyses.

Required packages

The following packages are needed to reproduce this vignette:

library(eurolig)
library(dplyr)
library(ggplot2)

Getting data

Datasets

Several datasets are included in the package. You can see all the available datasets in the package by calling:

data(package = "eurolig")

Sample datasets of play-by-play and shot location data are stored in samplepbp and sampleshots, respectively.

A particularly useful dataset in the package is gameresults. It contains all game results in the Euroleague from the 2001/2002 season to the 2018/2019 season. As you will see in the next section, this dataset can be useful to find the games you want to get data from.

Finally, if we need to find the name or identifying code of teams, the teaminfo dataset can be helpful.

extract functions

Functions that retrieve data from Euroleague's website API start with the verb extract. You will need to be online for these functions to work.

The main functions to get data are:

• extractPbp() for play-by-play data.

• extractShots() for shot location data.

These functions can only retrieve data from a single game. That means that if you want to get data for several games you will need to iterate the function over the games of interest. Note, however, that Euroleague's robot.txt asks for a 15 (!) seconds delay between requests. Take this into consideration when requesting data for a lot of games.

Games are uniquely identified by a combination of season and game code. In order to indicate the extract functions what game we want to get data from, we need to pass the corresponding game code and season as arguments.

Let's find the highest scoring games from the 2018/2019 season in the gameresults dataset:

games <- gameresults %>% 
  filter(season == 2018) %>% 
  mutate(total_points = points_home + points_away) %>% 
  arrange(desc(total_points)) %>% 
  head()

We can see that last season's highest scoring game was the Herbalife Gran Canaria vs. AX Armani Exchange Olimpia Milan with 210 total points scored between the two teams. The identifying game code and season for this game are

games$season[1]
games$game_code[1]

We can get play-by-play and shot location data for this game by passing these values as arguments to extractShots() and extractPbp():

game_shots <- extractShots(game_code = 168, season = 2018)
game_pbp <- extractPbp(168, 2018)

If you want to find games for the current season (not included in the gameresults dataset), you have two options: either look up the game code in the game's url or use extractResults().

Analyzing play-by-play data

Play-by-play data provides a lot of information that traditional boxscore statistics fail to communicate. In the following subsections I am going to show how we can find the following information from play-by-play data:

• Plus-Minus for one or more players.

• On/Off statistics for one or more players.

• Assists patterns within a team

For these analyses I am going to use the sample dataset samplepbp which contains the play-by-play data for all four games of the 2018/2019 Euroleague Final Four:

data("samplepbp")
glimpse(samplepbp)

Plus-Minus

Plus-minus (+/-) measures the difference in team points scored and team points allowed while a player or a set of players of the same team are on the court. getPlusMinus() parses a play-by-play data frame with one or more games and returns the indicated player/s plus-minus statistic in each game.

Although widely used nowadays, it is important to note that raw plus-minus is very unstable and totally context dependent. It is, however, the building block for other more advanced stats such as RAPM or RPM.

Let's check what Sergio Rodriguez (aka El Chacho) plus-minus was in the two Final Four games that he played in the 2018/2019 season:

chacho_pm <- getPlusMinus(pbp = samplepbp, players = "RODRIGUEZ, SERGIO")
# Select only a few columns so that data frame fits in the document
chacho_pm %>% 
  select(game_code, team_code_opp, poss, poss_opp, plus_minus)

Note that you can find the plus-minus statistic for combinations of players by entering a character vector with the player names in the players argument.

On/Off Statistics

On/Off statistics for a player or a set of players measure team statistics when the player or players where on the court and when they were on the bench.

You can use getOnOffStats() to find on/off statistics. For instance, I can find out how Real Madrid did when Rudy and Ayón were together on the court versus when both were on the bench in the two games played in the 2018/2019 Final Four.

getOnOffStats(pbp = samplepbp, players = c("FERNANDEZ, RUDY", "AYON, GUSTAVO"))

Note that getOnOffStats() returns 4 rows per game corresponding to:

• Team statistics when players were on the court together.

• Opposing team statistics when players were on the court together.

• Team statistics when players were on the bench together.

• Opposing team statistics when players were on the bench together.

Assists patterns

Play-by-play data allows us to find out who assists who on assisted baskets. The function getAssists() returns a data frame with the passer and the shooter (plus more contextual information) for all assisted baskets in the input play-by-play data.

I am going to use getAssists() and some data wrangling to find out the most common assists in CSKA Moscow during the two games of the 2018/2019 Final Four:

assists_csk <- getAssists(pbp = samplepbp, team = "CSK")
assists_csk %>% 
  count(passer, shooter) %>% 
  arrange(desc(n))

Analyzing shot location data

Shot location data specifies the x and y coordinates of every jump shot taken during a game. This data can be useful for, say, identifying shooting location tenedencies on offense of a given player/team, showing what spots on the court a player is most effective at, or analyzing what type of shots a team allows its opponents.

For the following analysis, I am going to use the sampleshots dataset, which contains the shot location data for the four games in the 2018/2019 Final Four.

glimpse(sampleshots)

You can see in addition to the x-y coordinates, coord_x and coord_y, there are variables that give you more contextual information such as the player that took the shot, the time in the clock when the shot was taken or whether the shot was after an offensive rebound. These variables can help you filter specific shot types.

The function plotShotChart() allows you to show graphically where the shots were taken on the court and colored according to whether the shot went in (green) or not (red). The returned object is a ggplot object that we can customize with ggplot2 functions:

plotShotchart(sampleshots) +
  labs(title = "Euroleague 2018/2019 Final Four") +
  theme(legend.position = "bottom")

solmos/euroleaguer documentation built on Jan. 30, 2022, 3:16 p.m.