Comparing Liverpool’s Full Backs using StatsBomb’s Radar Chart
With a little dash of data science.
Over the past two years, Liverpool have reaped success and having a world-class full-back tandem is one of the main factors amongst many. 19th league titles and 6th European titles come from Jürgen Klopp’s very own heavy metal mix and it involves chance creation from both sides of the field.
Trent Alexander-Arnold and Andrew Robertson played an integral part as Liverpool would like to build their attack from the flanks. Both players come second and third in the prem’s assists tally. Trent is second with 13, and Andy coming right behind with 12, playing for a cross-happy team that amassed 885 crosses in total, ranked second behind Manchester United’s 919.
Certainly there are other aspects in a full-back’s game and as a fan, I can’t help but wonder how they fare in those features? Since both players have a friendly competition going on between them, I’d like to spice it up by making a radar chart comparison inspired by StatsBomb.
The idea is to create a chart like this:
And in order to achieve that, I will be using PySport’s soccerplots Python library to visualize the data that I have retrieved from fbref.com. Let’s begin.
Data Retrieving
In this analysis, I’ll be using FBRef’s stats of Liverpool. There are six datasets included, which are:
- Passing Stats (passing),
- Goal and Shot Creation Stats (scagca),
- Defensive Actions Stats (defensive),
- Possession Stats (possession), and
- Playing Time Stats (time).
I downloaded each of them and created a schema database so I can query my data conveniently. I could’ve used Pandas to analyze and select the data I want to display, but I’ll stick SQL because it’s easier and I’d like to sharpen up my querying skill.
I believe fbref can’t provide us all the data displayed in the chart so let’s make the most out of what we have by displaying per 90 minutes stats of Expected Assist (xA), Key Passes (KP), Passes into Penalty Area (PPA), Crosses into Penalty Area (CrsPA), Deep Passes Completion (LCmp), Progressive Passes (Prog), Shot-Creating Actions (SCA), Goal-Creating Action (GCA), Dribblers Tackled Percentage(TklP), Defensive Actions (DA, adding Tackles, Interceptions, and Clearances), Successful Dribble Percentage (DrbP), and Expected Goals Plus-Minus (xGPM90).
Using multiple joins I retrieved data from five different tables and now it’s ready to process. Now I have to read the data by simply using Pandas before finally visualizing it…
With all being set, let’s get our hands dirty! I will not be explaining the know-hows of the code since it’s well beyond the scope of this article, but if you want to learn more about Python
## setting parameters
params = df.columns.tolist()## setting range values
ranges = [(0.15, 0.35), (1.00, 3.00), (1.50, 3.00), (0.50, 1.00), (10.0, 25.0), (5.00, 15.00),
(2.50, 5.00), (0.0, 1.00), (30.0, 50.00), (5.00, 6.00),
(40.00, 55.00), (0.70, 1.00)]## setting parameter value
values = df.iloc[0]## titles to make it pretty
title = dict(
title_name='Trent Alexander-Arnold',
title_color='#000000',
subtitle_name='Liverpool FC',
subtitle_color='#D00027',
title_fontsize=18,
subtitle_fontsize=15,
)## plotting the radar chart
fig, ax = radar_chart.plot_radar(ranges=ranges, params=params, values=values, radar_color=['#D00027', '#FFFFFF'], title=title)
And the result…
With the same method, we can plot the chart for Andy…
But wouldn’t it be easier for us to compare if we put two charts together? by adding both players' data as the values and compare=True command, we can have one chart overlaps the other so we can compare both.
## For Comparison. No need to add new ranges and params since it will be the same.
## setting parameter values of both players
val_comp = df.iloc[0], df.iloc[1]## titles for each players
title_comp = dict(
title_name='Trent Alexander-Arnold',
title_color='#D00027',
subtitle_name='Liverpool',
subtitle_color='#000000',
title_name_2='Andrew Robertson',
title_color_2='#00A398',
subtitle_name_2='Liverpool',
subtitle_color_2='#000000',
title_fontsize=18,
subtitle_fontsize=15,
)## plotting the radar chart
fig, ax = radar_chart.plot_radar(ranges=ranges, params=params, values=val_comp, radar_color=['#D00027', '#00A398'], title=title_comp, compare=True)
And the result…
From the chart, it appears that Alexander-Arnold was superior on many playmaking aspects compared to Robertson. It supports Michael Cox’s analysis of the Englishman stating that he’s not a mere full-back. Alexander-Arnold play lofted passes, quick one-twos, and also cutting inside more often than your usual wide defender outlets, making him the prime playmaker of Liverpool.
While not being on the same level in terms of playmaking, Robertson brought values from his defending skills as he’s better in tackling and also dribbling the ball. You can say that he’s a classic full-back, who might not bring new dimensions in his game but he excels in the fundamentals. And surprisingly, he brought more values in goalscoring when he’s on the pitch as shown from his expected goals plus-minus number.
Of course, analyzing football players and comparing them needs more than just numbers written on the sheets and displayed by a chart, But it’ll definitely help you to get a good picture of how they’re performing on the pitch.
Data sources: FBref
Visualization method: Slothfulwave612
PySport’s website: here
This article’s Python notebook: Available on my GitHub Repo
