Abhijit Bharali | Twitter: @abhibharali
This is a recreation of a dataviz used by Stats Perform in their cricket data science related projects. The visualisation, similar to visual representation of time series data, showcases how batters pace their innings in a match, highlighting that not all runs are scored similarly. View my full project documentation here.
The viz goes like...
A binmap scatter is useful to highlight key details of a football match. It can be remodelled to suit various match-level contexts and visualise plenty of information by mapping pertinent data points from match event data. The full project documentation.
The viz goes like...
*** Method 1 ***
Copy-paste the data from transfermarkt.com in local machine and generate the viz.
*** Method 2 *** ⬇️ ⬇️ tutorial⬇️ ⬇️
Web scrape the data from transfermarkt.com and do a full-stack approach to building the viz.
I walk through method #2 here as it is computationally more challenging.
The transfermarkt.com link for the results to visualise is this, a complete list of results in the 2019-20 Indian Super League league phase.
The page shows like this...
We use the rvest package to scrape text from the page. For visualisation, we use the ggplot2 package. Other packages in use are magrittr for piping, stringr for text cleaning and extrafont for using custom fonts.
# Scrape the page and store it to a variable (sp in this example)
sp <- read_html("https://www.transfermarkt.co.in/indian-super-league/gesamtspielplan/wettbewerb/IND1?saison_id=2019")
# Generate a dataframe of the scraped data and store it to a variable (league in this example)
league <- sp%>%html_nodes(".hauptlink , .hauptlink .tooltipstered , td.hide-for-small:nth-child(1)")%>%
html_text()%>%
str_replace_all("[\t\n]", "")%>%
as.data.frame()
# Clean the league dataframe and convert it to a readable format (table dataframe in this example)
colnames(league)="date"
league$home<-league$date
league$result<-league$home
league$away<-league$home
for (i in 0:89) {
league$date[4*i+2]=""
league$date[4*i+3]=""
league$date[4*i+4]=""
league$home[4*i+1]=""
league$home[4*i+3]=""
league$home[4*i+4]=""
league$result[4*i+1]=""
league$result[4*i+2]=""
league$result[4*i+4]=""
league$away[4*i+1]=""
league$away[4*i+2]=""
league$away[4*i+3]=""
}
table<-data.frame(
date=(league%>%filter(date!=""))$date,
home=(league%>%filter(home!=""))$home,
result=(league%>%filter(result!=""))$result,
away=(league%>%filter(away!=""))$away
)
table<-table%>%
separate(col = date,into = c("day","date"),sep = " ")
table$result2<-table$result
table<-table%>%
separate(result2,c("hgoal","agoal"),sep = ":")
table$home<-gsub('([(]+[0-9]+[.]+[)])','',table$home)
table$away<-gsub('([(]+[0-9]+[.]+[)])','',table$away)
# remove unwanted trailing & leading whitespaces in the table dataframe
table$home<-str_trim(table$home,side = "both")
table$away<-str_trim(table$away,side = "both")
table$date<-str_trim(table$date,side = "both")
# convert hgoal & agoal to numbers (first convert to factors to override NA coercion error)
table$hgoal<-as.factor(table$hgoal)
table$hgoal<-as.numeric(table$hgoal)
table$agoal<-as.factor(table$agoal)
table$agoal<-as.numeric(table$agoal)
table$hgoal <- table$hgoal-1
table$agoal <- table$agoal-1
glimpse(table)
head(table%>%as_tibble())
# Visualise the data and save the plot to local machine using `ggsave`
ggplot(table,aes(x=away,y=home,fill=hgoal-agoal))+
geom_tile(alpha=0.5)+
geom_text(aes(label=paste(hgoal,agoal,sep = "-"),
family="Helvetica",size=hgoal+agoal),
colour="black",
show.legend = F)+
scale_fill_gradient2(low = "red",high = "green",midpoint = 0,guide = FALSE)+
scale_x_discrete(limits = levels(table$home), position = "top") +
scale_y_discrete(limits = rev(levels(table$away)))+
labs(title="ISL 19-20 RESULT MATRIX",
caption="darker green = more emphatic home win\ndarker red = more emphatic away win")+
xlab("AWAY")+
ylab("HOME")+
theme(text = element_text(family = "Helvetica"),
axis.ticks = element_blank(),
panel.grid = element_blank(),
panel.background = element_rect(fill = "white"),
plot.background = element_rect(fill = "white"),
plot.title = element_text(size = 18,hjust = 0,family = "Helvetica",
colour = "midnight blue",face = "bold"),
plot.caption = element_text(hjust = 0.5),
axis.text = element_text(size = 7))+
ggsave("result matrix",dpi = 300, device = "png",width = 11.27,height = 8.27)
The final result looks like this...
This is an example of a full-stack sports data visualisation approach using R, wherein we extract the data from the web using scraping techniques, clean and wrangle the extracted data to a usable format and then visualise the data to generate insights.
One of the most important aspects of performance analysis in any sport is opposition analysis. This lollipop chart visualisation provides the starting point for opposition analysis for football teams, highlighting the changes made to a team's starting XI on a per-game basis.
Here's the source code for automating lineup information using the worldfootballR package for extracting the required data and ggplot2 for visual representation and analysis.
The end product looks like this...
A widely used visualisation tool to compare players and teams. Its creation is aided by the fmsb package. View my full project documentation here.
The viz goes like...
This full-stack application makes use of rich ball-by-ball data of every Indian Premier League (IPL) match in history to generate in-depth insights. This project encompasses web scraping, data wrangling, mining, modelling and visualisation, a full package . Project source code here.
A little preview of the application...
Untitled.mp4
An attempt to apply a logistic regression classifier to passes in football to value them and predict their success.
A little preview of the application...