xG Accumulative Charts

Expected Goals (xG) is a popularised metric for calculating the probabilty that a particular shot will be scored - more details. xG is a more reliable metric when assessed over seasons worth of data and single-game analysis using xG has huge weaknesses. However, single-game accumulative analysis has become popular, mainly because it shows a narrative of the match over time.

Using Statsbomb free women's football data I will run through the code to create the xG Accumulative plots.

The Data & Setup

I will be using the recently released dataset from Statsbomb, which is free to use and easy to access. The dataset consists of 11 games at the time of writing, which is equivilent to over 25,000 events.

Data Load

events <- readRDS("SB_events_DB.RDS") ## an events dataframe loaded from local storage - see my previous tutorial 

Package Loading

require(dplyr); require(ggplot2)

## Loading required package: dplyr

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

## Loading required package: ggplot2

Filtering to one match and two teams

## pick a random match from the Statsbomb dataset 
randomMatch <- unique(events$match_id)[runif(1, 1, length(unique(events$match_id)) )]

## select just the shots from that match 
shotData <- events %>% filter(match_id == randomMatch & shot.statsbomb_xg > 0)

## create a more precise metric for 'time' which combines minutes and seconds 
shotData$Timer <- shotData$minute + (shotData$second / 60)

## drop all unneeed data to make it easier to work with the data 
shotData <- shotData[c(19,54,61,147)]

## Create a dataframe just for the Team 1 and Team 2
T1_shotData <- shotData %>% filter(team.name == unique(shotData$team.name)[1])
T2_shotData <- shotData %>% filter(team.name == unique(shotData$team.name)[2])

Create the accumulative values for xG for each team

We calculate the accumulative xG total using the cumsum() function which comes pre-installed with R install. This gives us back just the data we need.

## Team 1
# Calculate the accumulative xG total and add it as a column to the dataframe 
T1_shotData$xg_total <- cumsum(T1_shotData$shot.statsbomb_xg)
# create the starting and ending rows of data to ensure that all teams start on 0 and end at 96 minutes 
Top2add <- data.frame(team.name = unique(shotData$team.name)[1], shot.statsbomb_xg = 0, Timer = 0, shot.outcome.name = "-", xg_total = 0, stringsAsFactors = F)
Bottom2add <- data.frame(team.name = unique(shotData$team.name)[1], shot.statsbomb_xg = 0, Timer = 96,  shot.outcome.name = "-", xg_total = max(T1_shotData$xg_total), stringsAsFactors = F)

## add these rows to the data with bind_rows()
T1_shotData <- bind_rows(T1_shotData, Top2add)
T1_shotData <- bind_rows(T1_shotData, Bottom2add)


## Team 2
# Calculate the accumulative xG total and add it as a column to the dataframe 
T2_shotData$xg_total <- cumsum(T2_shotData$shot.statsbomb_xg)
# create the starting and ending rows of data to ensure that all teams start on 0 and end at 96 minutes 
Top2add <- data.frame(team.name = unique(shotData$team.name)[2], shot.statsbomb_xg = 0, Timer = 0,  shot.outcome.name = "-", xg_total = 0, stringsAsFactors = F)
Bottom2add <- data.frame(team.name = unique(shotData$team.name)[2], shot.statsbomb_xg = 0, Timer = 96,  shot.outcome.name = "-", xg_total = max(T2_shotData$xg_total), stringsAsFactors = F)

## add these rows to the data with bind_rows()
T2_shotData <- bind_rows(T2_shotData, Top2add)
T2_shotData <- bind_rows(T2_shotData, Bottom2add)

Plot the Data

Now we have all the data to plot lets build our plots, as always I build it layer by layer to help learning. We will use the trusted ggplot2() package.

We will use the geom_step() function that is designed specifically for the type of chart we want and takes all the hard work out of it!

Basic Plot

p <- ggplot()
p

Add Team 1 Data

p <- p + geom_step(data=T1_shotData, mapping=aes(x=Timer, y=xg_total), colour = "#E24F55", size = 1.5, alpha = 0.8)
p

Add Team 2 Data

p <- p + geom_step(data=T2_shotData, mapping=aes(x=Timer, y=xg_total), colour = "#2B6DD2", size = 1.5, alpha = 0.8)
p

Add the Goals

### prepare the data 
T1_goals <- T1_shotData %>% filter(shot.outcome.name == "Goal")
T2_goals <- T2_shotData %>% filter(shot.outcome.name == "Goal")

## add the points on for the goals 
p <- p + 
geom_point(data = T1_goals, aes(x=Timer, y=xg_total), colour = "black", size = 4, alpha = 1, fill = "#E24F55", shape = 21) +
geom_point(data = T2_goals, aes(x=Timer, y=xg_total), colour = "black", size = 4, alpha = 1, fill = "#2B6DD2", shape = 21)

p

Labels, Theme & Legends

## find the axis length for the plot
yMax <- ggplot_build(p)$layout$panel_ranges[[1]]$y.range

## create the annotation titles 
T1_label <- paste0(unique(shotData$team.name)[1], ": ", nrow(T1_goals), " Actual: ", round(max(T1_shotData$xg_total), 2), "xG Diff: ", nrow(T1_goals) - round(max(T1_shotData$xg_total), 2))

T2_label <- paste0(unique(shotData$team.name)[2], ": ", nrow(T2_goals), " Actual: ", round(max(T2_shotData$xg_total), 2), "xG Diff: ", nrow(T2_goals) - round(max(T2_shotData$xg_total), 2))

## add the legend 
p + 
annotate("text", x = 0, y = (yMax[2] - (yMax[2]*0.05)), label = T1_label, hjust = 0, colour = "#E24F55") + 
annotate("text", x = 0, y = (yMax[2] - (yMax[2]*0.1)), label = T2_label, hjust = 0, colour = "#2B6DD2") + labs(x = "Minutes", y = "Accumulative xG") + theme_minimal()

There are many ways to style these plots but the above code will get you to the final product for you to tweak further.. enjoy.