5 days ago, Sky
Sports posted this image on Twitter. Among the soccer/football analytics
community, this was excellent to see given the usage of Sky Sports of Expected
Goals - a metric which has been actively pushed amomg bloggers, and indeed also
used within clubs.
While enthused by
the progess, I was less than excited by the visualisation which Sky Sports used
to display this metric, among others. It took a while to compute what was going
on. Let's break it down…
So the above
graphic's objective is to investigate Arsene Wenger's claim that Lacazette is a
good finisher. So it compares 3 metrics, and minutes played as a contextual
figure among 7 of Arsenal's forwards.
Now, I'm a fan of
donut charts (well, two segment donuts with a number in the middle as shown
here) - however, the way it's used above is misleading. Let me explain why. It
took me a while to compute, but the 'blue' full donuts are highlighting the
highest among the 7 players, and the 'completed' segment of the donut (ie. The
red part) how close this is to the highest in that metric.
While it took a
while for me to understand that, it's not overly intuitive. I saw a really nice
summary from Steve Wexlar which said
"The goal for your viz: allow the largest number of people to get the
greatest degree of understanding with a minimum amount of effort." - Given
that it took longer than expected to understand the chart, I don't think this
is the optimal chart for the message that Sky Sports want to convey.
So what can be done?
After a couple of
tweets with Mark (@everyteam_mark) and Danny (@DannyPage), I hopped into Tableau
and built out the following examples as alternatives to the visualisation.
The most basic ways
I thought this could be re-done was using dots and bars. The bar chart is the
most intuitive way of viewing data as the length is easy to compare - however,
it's deathly boring and super functional. A dot plot is similar to a bar, but
instead of encoding length, it uses position along an axis. Instead of an axis
here, I've used labels so it's easier to attribute a value to the mark.
An immediate
alternative which came to mind was a bump chart - this is effectively a way
rank a player along the four metrics. However, this does mean that the detail
is lost within the raw difference between each player as it only highlights the
order and ranking. It does, however show that for all the metrics, that
Lacazette is second - which supports the claim the article was making - but as
mentioned before, the detail is lost.
My personal
favourite alternative was this one - it effectively pivotes the dot plot above,
but as (at least in the western world) we read left to right, it's easier to
compare the position of the dots, and Tableau also allows us to highlight two
players at once. It also shows rank, comparison and also raw difference between
the players.
These are just a few
examples of how I'd quickly do a makeover of what Sky Sports showed - I also
really liked how Simon (@analytic_footy) showed an alternative to the bump
chart, highlighting Lacazette in red, with the others greyed out; which was a
great touch and added some great context. I am definitely a fan.
How would you
visualise this data?
Find it below, and
have a go - and let me know how you get on!
Player
|
Mins
played
|
Shot
conversion %
|
Goals
per 90 (exc. Pens)
|
xG per
90 (exc. Pens)
|
Lacazette
|
2,408
|
33.3%
|
0.67
|
0.47
|
Sanchez
|
3,223
|
18.6%
|
0.61
|
0.42
|
Giroud
|
1,194
|
30.8%
|
0.9
|
0.51
|
Walcott
|
1,925
|
16.1%
|
0.47
|
0.36
|
Welbeck
|
746
|
10.5%
|
0.24
|
0.38
|
Perez
|
265
|
11.1%
|
0.34
|
0.27
|
Iwobi
|
1,462
|
7.69%
|
0.18
|
0.18
|
0 comments:
Post a Comment