Yesterday I tweeted about a confusing graphic from a BBC article dispelling myths about last season's Premier League. Its a good article with some really interesting content but some of the accompanying graphics hindered rather than enlightened the points being made.
How to make something unnecessarily hard to read, dotty-donut edition http://t.co/OKBo3ZdFDt pic.twitter.com/NnNKIX67Lh
— Andy Kirk (@visualisingdata) August 12, 2015
A couple of people asked what I would do instead. Time rarely allows it but one should always be willing to offer an alternative, even if it is just quickly expressing some ideas verbally. Criticism without suggestion is empty and it is something I fall into all too often. So I thought I would quickly offer a reworking of the two main graphics that I felt caused unnecessary inaccessibility. I assumed the same constraints around space, (similar) typeface and colours as well as the inclusion of the logo and hashtag.
The first concerns the use of a radar chart to demonstrate how Stoke City have evolved from a long-ball team under Tony Pulis, into a more progressive team under Mark Hughes. With Radar Charts, there are almost always better solutions, especially when you are attempting to compare two series of values in the same chart. Radars really only make sense if and when there is some compelling logic for the radial arrangement of values (usually temporal, spatial or, occasionally, intuitive groups).
The redesign uses a connected dot plot to draw out the differences in rankings between the different measures. The measures themselves are re-ordered to try and group related ones together.
The second relates to the main graphic I mentioned on Twitter, which looks at dispelling the idea that teams with the most possession have the most success. The donut chart was reasonably fine until the dots landed. Due to their placement within the arcs there is an implication of meaning, especially when we see the two adjacent dots for the call out aggregating the loss and drawn match percentages. Unnecessarily confusing when all we need to learn about is 4 numbers.
The redesign is very simply a stacked bar-chart. To be honest it still doesn't add loads of value as a visual, you are essentially getting most of your understanding from the value captions but the stacked bar better aids showing the 'did not win' aggregate. I've switched the colours to perhaps better suggest bad, medium, good.
Comparing the before/after versions I suspect that the labelling size and prominence of colour of my redesigns would need fine-tuning. It is interesting to see how faded they look when you shrink the final png file down. In the native Illustrator version they look far more vivid to the naked eye. That's a good lesson in testing out your prototype designs in the size and setting in which they are likely to exist, to see for yourself how they look. Anyway, I've not got time to undertake endless iterations but you get the idea.
Congratulations. You made a very clear improvement in both graphs. In the first, more complex reading data, a pattern can be seen very clearly in the improved graphics, which is not seen in the “radar chart”: the team under Pulis was more radical in their records, stood at the ends in the Premier League ranking. Team under Hugues, the results are more intermediate positions.
However, I have a question about this chart “connected dot plot”: Why in the “X” axis starts at 20 and ends at 1? Is there any reason why you cann’t go from 1 to 20? Thanks.
Thanks Francesc! Yes, the x value sequences was a flip-of-a-coin decision, could have gone either way. My final decision was based on the idea that we would normally read, for example, a horizontal bar chart going more to the right as being ‘higher’ so I used that in terms of a higher ranking being reached. In this case it is not necessarily always ‘good’ to be higher ranked (more fouls is not a ranking you particularly want to achieve) so I didn’t try to imply higher is better, just more to the right is higher in ranking.
Hi Andy. Those BBC charts are poorly thought out aren’t they.
My first thought on the first graphic would be to use a SlopeGraph approach with each line being an attribute and the y-axis going from 1st to 20th. This way the line gradient allows the user to interpret the direction and magnitude of the change very quickly.
Hi David. Yes, I considered the slope as my first option because as you say its great at showing up/down changes. Two things undermined it working as well as I hoped. Firstly, the rectangular landscape area to work wasn’t best for occupying what are normally quite portrait chart areas. Secondly, because there were a number of repeated ranking positions across the metrics, the slope labelling and line overlapping would have been rather busy.
[…] Quick Redesigns of BBC Sport Graphics […]
[…] 원문보기 […]
I’m a neophyte in the world of data visualization (though I greatly enjoyed one of Tufte’s classes at Uni in the late 80’s). Just this week I published a blog post on my use of the Radar Chart in my tennis statistics application (since updated to link to this piece). I largely agree with your (and others) criticisms of Radar Charts. You listed a few situations (temporal, spatial, intuitive groups) where the case for radial charts could be compelling and I wondered what you might think about the scenario I present where an attempt is made to use an iconic (no labels) version to quickly compare a series of matches across a group of key statistics. There is also the difference that in an application the graphic is used interactively and repeatedly; it is not a one-off for a research paper or work of journalism…
http://tennisviz.blogspot.com/2015/08/match-radar.html