Homework 2

hw2

Prasann Desai

worldfootballR

Reading in Data

Author

Prasann Desai

Published

July 3, 2023

1 Introduction

Until a decade and a half ago, recording data points regarding key events in soccer was a manual process. This limited the range of analysis and the variety of data attributes. Today, with the usage of advanced technology, hundreds of attributes describing a game are being recorded in real-time. This has enabled a wide range of research in the field of soccer analytics and led to the emergence of various soccer data aggregators like StatsBomb, FBref, Understat, Fotmob, etc. In this project, we will be using the data from these aggregators using the worldfootballR package in R. This package has been created by Jason Zivkovic that enables users to conveniently access data belonging to the above-mentioned aggregators in R.

The goal of the project will be to find performance attributes that best predict success in the game. We would want to see the following aspects:

What are the best teams/players good at? We will compare their aggregate statistics like goals, assists, xG, xA, Goal difference, etc.
Is xG a good indicator of success? Any trends over years?
Which team has been the most unlucky? Here, we will analyse the difference between XG and Actual goals to see which teams couldn’t secure points despite performing well.

2 Data

This dataset allows us to obtain soccer game/team/player data with various attributes. The attributes are both categorical and numerical. The attributes describing a game are goals scored by the home/away team, assists, game result, corners taken, shots attempted, etc. Similar attributes are also recorded at a team and player level. The dataset also contains some advanced attributes like xG and xA that are widely accepted in the soccer analytics industry to describe the quality of actions of players/teams in quantitative terms. We will be making use of xG and related stats to determine if it is a good indicator of success (better points earned in the league, more goals scored, etc.). For simplicity, we will be using the stats of the 1st tier of English Premier League Football for our analysis.

# This is a sample dataset call. Further in the project, we could change a couple of parameters in the dataset call for comparative analysis, etc.
prem_team_2023_shooting <- fb_season_team_stats(country = "ENG", gender = "M", season_end_year = "2023", tier = "1st", stat_type = "shooting")

nrow(prem_team_2023_shooting)

[1] 40

table(prem_team_2023_shooting$Team_or_Opponent)


opponent     team 
      20       20

Attribute Information

Competition_Name (str, categorical): Name of the Competition
Gender (str, categorical): Gender category of the competition
Country (str, categorical): Country of the competition
Season_End_Year (int, numerical): Year of end of the competition season
Squad (str, categorical): Name of the team
Team_or_Opponent (str, categorical): Takes only two values - ‘team’ and ‘opponent’. If the record is a ‘team’ record, then the stats are scored by the Squad, else, the stats are scored against the Squad.
Num_Players (int, numerical): Number of players registered in that season
Mins_Per_90 (int, numerical): Games played per season
Gls_Standard (int, numerical): Goals scored in a season
Sh_Standard (int, numerical): Shots taken in a season
SoT_Standard (int, numerical): Shots on target in a season
SoT_percent_Standard (int, numerical): Ratio of SoT_Standard/Sh_Standard
Sh_per_90_Standard (int, numerical): Sh_Standard per 90 mins
SoT_per_90_Standard (int, numerical): SoT_Standard per 90 mins
G_per_Sh_Standard (int, numerical): Gls_Standard/Sh_Standard
G_per_SoT_Standard (int, numerical): Gls_Standard/SoT_Standard
Dist_Standard (int, numerical): Distance travelled by the team in the season. (Running distance)
FK_Standard (int, numerical): Free kicks taken in a season
PK_Standard (int, numerical): Penalty kicks scored in a season
PKatt_Standard (int, numerical): Penalty kicks awarded in a season
xG_Expected (int, numerical): Expected Goals scored in a season
npxG_Expected (int, numerical): non-penalty Expected Goals scored in a season
npxG_per_Sh_Expected (int, numerical): non-penalty Expected Goals per shot in a season
G_minus_xG_Expected (int, numerical): Difference between actual goals scored and Expected goals scored
np:G_minus_xG_Expected (int, numerical): Difference between actual non-penalty goals scored and non-penalty Expected goals scored

# Sample analysis
# Pre-processing
# selecting the required columns
prem_2023_shooting_1 <- select(prem_2023_shooting, Squad, Team_or_Opponent, npxG_Expected)

Error in select(prem_2023_shooting, Squad, Team_or_Opponent, npxG_Expected): object 'prem_2023_shooting' not found

prem_2023_shooting_1 <- mutate(prem_2023_shooting_1, Squad = case_when(str_detect(Squad, "vs ") ~ str_split(Squad, "vs ", simplify = TRUE)[,2],
                                                                     TRUE ~ as.character(Squad)
                                                                     ))

Error in mutate(prem_2023_shooting_1, Squad = case_when(str_detect(Squad, : object 'prem_2023_shooting_1' not found

# Using the pivot_wider function to pivot the dataset
prem_2023_shooting_pivoted <- pivot_wider(prem_2023_shooting_1, names_from = "Team_or_Opponent", values_from = "npxG_Expected")

Error in pivot_wider(prem_2023_shooting_1, names_from = "Team_or_Opponent", : object 'prem_2023_shooting_1' not found

prem_2023_shooting_pivoted <- rename(prem_2023_shooting_pivoted, xG_For = team, xG_Against = opponent)

Error in rename(prem_2023_shooting_pivoted, xG_For = team, xG_Against = opponent): object 'prem_2023_shooting_pivoted' not found

prem_2023_shooting_pivoted

Error in eval(expr, envir, enclos): object 'prem_2023_shooting_pivoted' not found

# Building a scatter plot of xG_For and xG_Against

ggplot(prem_2023_shooting_pivoted,
       aes(x = xG_For,
           y = xG_Against,
           col = Squad
           )) +
  geom_text(aes(y = xG_Against - 1, label = Squad), size = 3) + 
  geom_point(size = 2.5, alpha = 0.8) + 
  labs(x = "Expected Goals Scored",
       y = "Expected Goals Conceded",
       title = "xG_For vs. xG_Against",
       subtitle = "xG For vs. Against comparison of all Premier league teams for season of 2022-2023") +
  theme(legend.position = "none")

Error in ggplot(prem_2023_shooting_pivoted, aes(x = xG_For, y = xG_Against, : object 'prem_2023_shooting_pivoted' not found

--- title: "Homework 2" author: "Prasann Desai" description: "Reading in Data" date: "7/3/2023" format: html: toc: true toc-depth: 2 toc-title: Contents toc-location: left number-sections: true number-depth: 2 code-copy: true code-tools: true categories: - hw2 - Prasann Desai - worldfootballR --- ```{r} #| label: setup #| warning: false #| message: false #| include: false library(tidyverse) library(purrr) library(ggplot2) knitr::opts_chunk$set(echo = TRUE, warning=FALSE, message=FALSE) # to allow us to install packages from github and eventually install/load up the dataset. # install.packages("devtools") # devtools::install_github("JaseZiv/worldfootballR", ref = "main") library(worldfootballR) ``` ## Introduction Until a decade and a half ago, recording data points regarding key events in soccer was a manual process. This limited the range of analysis and the variety of data attributes. Today, with the usage of advanced technology, hundreds of attributes describing a game are being recorded in real-time. This has enabled a wide range of research in the field of soccer analytics and led to the emergence of various soccer data aggregators like StatsBomb, FBref, Understat, Fotmob, etc. In this project, we will be using the data from these aggregators using the *worldfootballR* package in R. This package has been created by Jason Zivkovic that enables users to conveniently access data belonging to the above-mentioned aggregators in R. The goal of the project will be to find performance attributes that best predict success in the game. We would want to see the following aspects: - What are the best teams/players good at? We will compare their aggregate statistics like goals, assists, xG, xA, Goal difference, etc. - Is xG a good indicator of success? Any trends over years? - Which team has been the most unlucky? Here, we will analyse the difference between XG and Actual goals to see which teams couldn't secure points despite performing well. ## Data This dataset allows us to obtain soccer game/team/player data with various attributes. The attributes are both categorical and numerical. The attributes describing a game are goals scored by the home/away team, assists, game result, corners taken, shots attempted, etc. Similar attributes are also recorded at a team and player level. The dataset also contains some advanced attributes like xG and xA that are widely accepted in the soccer analytics industry to describe the quality of actions of players/teams in quantitative terms. We will be making use of xG and related stats to determine if it is a good indicator of success (better points earned in the league, more goals scored, etc.). For simplicity, we will be using the stats of the 1st tier of English Premier League Football for our analysis. ```{r} # This is a sample dataset call. Further in the project, we could change a couple of parameters in the dataset call for comparative analysis, etc. prem_team_2023_shooting <- fb_season_team_stats(country = "ENG", gender = "M", season_end_year = "2023", tier = "1st", stat_type = "shooting") ``` ```{r} nrow(prem_team_2023_shooting) ``` ```{r} table(prem_team_2023_shooting$Team_or_Opponent) ``` ### Attribute Information 1) Competition_Name (str, categorical): Name of the Competition 2) Gender (str, categorical): Gender category of the competition 3) Country (str, categorical): Country of the competition 4) Season_End_Year (int, numerical): Year of end of the competition season 5) Squad (str, categorical): Name of the team 6) Team_or_Opponent (str, categorical): Takes only two values - 'team' and 'opponent'. If the record is a 'team' record, then the stats are scored by the Squad, else, the stats are scored against the Squad. 7) Num_Players (int, numerical): Number of players registered in that season 8) Mins_Per_90 (int, numerical): Games played per season 9) Gls_Standard (int, numerical): Goals scored in a season 10) Sh_Standard (int, numerical): Shots taken in a season 11) SoT_Standard (int, numerical): Shots on target in a season 12) SoT_percent_Standard (int, numerical): Ratio of SoT_Standard/Sh_Standard 13) Sh_per_90_Standard (int, numerical): Sh_Standard per 90 mins 14) SoT_per_90_Standard (int, numerical): SoT_Standard per 90 mins 15) G_per_Sh_Standard (int, numerical): Gls_Standard/Sh_Standard 16) G_per_SoT_Standard (int, numerical): Gls_Standard/SoT_Standard 17) Dist_Standard (int, numerical): Distance travelled by the team in the season. (Running distance) 18) FK_Standard (int, numerical): Free kicks taken in a season 19) PK_Standard (int, numerical): Penalty kicks scored in a season 20) PKatt_Standard (int, numerical): Penalty kicks awarded in a season 21) xG_Expected (int, numerical): Expected Goals scored in a season 22) npxG_Expected (int, numerical): non-penalty Expected Goals scored in a season 23) npxG_per_Sh_Expected (int, numerical): non-penalty Expected Goals per shot in a season 24) G_minus_xG_Expected (int, numerical): Difference between actual goals scored and Expected goals scored 25) np:G_minus_xG_Expected (int, numerical): Difference between actual non-penalty goals scored and non-penalty Expected goals scored ```{r} # Sample analysis # Pre-processing # selecting the required columns prem_2023_shooting_1 <- select(prem_2023_shooting, Squad, Team_or_Opponent, npxG_Expected) prem_2023_shooting_1 <- mutate(prem_2023_shooting_1, Squad = case_when(str_detect(Squad, "vs ") ~ str_split(Squad, "vs ", simplify = TRUE)[,2], TRUE ~ as.character(Squad) )) # Using the pivot_wider function to pivot the dataset prem_2023_shooting_pivoted <- pivot_wider(prem_2023_shooting_1, names_from = "Team_or_Opponent", values_from = "npxG_Expected") prem_2023_shooting_pivoted <- rename(prem_2023_shooting_pivoted, xG_For = team, xG_Against = opponent) prem_2023_shooting_pivoted ``` ```{r fig.width = 10} # Building a scatter plot of xG_For and xG_Against ggplot(prem_2023_shooting_pivoted, aes(x = xG_For, y = xG_Against, col = Squad )) + geom_text(aes(y = xG_Against - 1, label = Squad), size = 3) + geom_point(size = 2.5, alpha = 0.8) + labs(x = "Expected Goals Scored", y = "Expected Goals Conceded", title = "xG_For vs. xG_Against", subtitle = "xG For vs. Against comparison of all Premier league teams for season of 2022-2023") + theme(legend.position = "none") ```