Background
For Day #2 of the Dashboard week challenge, we were presented with a dataset about English Women’s Football. The Ask – find insights from the data that are not mere overviews. Bascially, we had to dig for a story, not a dashboard with info like how many teams are there and how many goals did each team kick.
Discliamer: I know nothing about football. Let alone Women’s football. So when I looked at the dataset with field headers like tier, division, xxx I was lost. I had no idea what they meant. Some of the team names were easily recognisable. I guess even the non-footballers have heard of Arsenal! Since I don’t have domain knowledge of the dataset, I did not ask questions regarding team success and intricacies of the game. Instead I decided to try a different appraoch. I asked the question, “Has Women’s Football gained popularity over the years?”
For this I looked at number of people that attended the match. This resided in a field called “Attendance”. I decided to make that the focus of my data investigation. First I drew a simple line chart to plot the attendance numbers over time. It showed that there is indeed more people watching English Women’s football than there were 10 years ago.
Using a Violin Plot to See the Data Distribution
Then I drew a violin plot with a box and whisker plot. A violin plot can identify if there are more than one peak in thedistribution of your data. When I plotted the attendance per season, I could see that the past 4 seasons have had a signifiant jump in attendance. However, the jump is a greater jump in numbers, but the frequency of those numbers is not enough for it to be at the centre of the violin but it is beyond the 3rd quartile mark. So basically, the attendance is a huge jump but probably only for a few matches in a season, not all.
So I decided to drill down and explore these high but less frequent jumps and asked the question, “What affects attendance?”
What increases match attendance: the later matches in the season or the teams?
I hopothsized that it could either be the matches at the end of a season like the quarter-finals onwards, the matches that make or break, and the matches that are played by certain teams.
So I decided to tease this apart. First I created a new column through Dax that identified matches that were plotted in the 3rd quartile and above in the boxplots and called them “High Attendance” while the rest were “Low Attendance”. Then I created a scatter plot of the matches clubbed by the sequence they were played in and their attendance, coloured by High or Low Attendance.
Lastly I ranked the matches based on attendance and created a table visual with the top 50 ranks, as well as match sequence number, attendance catergory set to “High Attendance”, attendance numbers, names of both the teams and the season. I then highligted the matches that have a high attendance, but have occured early in the season.
To be honest, I did not use any of the cool features Power BI provides for interactivity, like drill downs or swapping charts etc. But my focus was more on deducing insights from the data; starting from a big picture place and narrowing it down to the details.
Conclusion
Nine out of the top 50 matches that had high attendance occurred early in the season, while the rest were played later. These nine matches were played by Arsenal Women, Chelsea Women, Manchester United Women and Aston Villa Women as home teams. This analysis shows that the matches that happen later in the season attract more crowds, but even the early season matches, if played by certain teams, attract large crowds too. Thus, teams that are playing can override the match sequence to attract crowds.
