Choosing a Chart - What Works, What Doesn't
The February data for Sports Viz Sunday looked at the Men & Women’s Six Nations Rugby Championships. The very comprehensive dataset included everything from the scores in each game, right down to the number of wooden spoon winners. And all of this going back to 1883!
Anyway, this is my visualisation. Click the image for the interactive version.
This was something that I could have kept adding in extra info everywhere: more detail in the tooltips, summaries of each team, the final standings of each championship. But sometimes you’ve got to call it day, otherwise you’ll never finish anything. It was also a visualisation I procrastinated over. After exploring the data at the start of February and having an overall layout and theme already built, I left it untouched for nearly two weeks. If this happens with a project of mine it usually spells doom, and it’s unlikely ever to see the light of day.
This viz-block was caused by the “main” chart in the viz; I just couldn’t work out how I should visualise the number of games won per nation, per year. The main thing I wanted was to make comparisons between years easy. I’d already decided to have a horizontal timeline running across the centre of a landscape dashboard. This worked really well with the data, especially as the Women’s competition started so much later than the Men’s, this allowed space for the title and various other bits & bobs in the top left.
Here’s the rough draft of the layout
In my mind this limited the ways I could visualise the data. To fit into the space set aside for the Women’s & Men’s competition, I needed something that worked horizontally. So after some playing with the data, here are four examples that I came up with, you can find this in the interactive version by clicking the image below.
You might notice that none of these ended up in the final viz, but doing this sort of thing really helps to get the me thinking about what works, and what doesn’t.
Version 1 - What Works
The countries that got 5 wins are really clear. The intention of using squares that “fill” from right to left was to give a bar chart effect when looking vertically.
It sort of merges into one long barcode, there just isn’t enough horizontal space on the dashboard to do it justice. Interesting enough, but I’m struggling to tell what’s going on - nope.
Version 2 - What Works
Aesthetically, this is my favourite, the curves and changing size of each line look really good.
Comparisons are difficult. It’s really hard to see where each year is, there’s definitely some overlapping going on. And no matter what I tried, I just couldn’t get rid of the line across the two world wars, just for that it’s an automatic fail.
Version 3 - What Works
Old reliable. The workhorse. It’s a bar chart and you can’t really go wrong with one of these. Yearly comparisons are nice and clear. Setting the y-axis to 5 highlights countries that win all their games; if the bar hits the bar above it, they won five games.
In regard to this viz, Its strength is also its weakness. Bar charts are found everywhere and I want something a little bit different to draw in the reader.
Version 4 - What Works
Very similar to version 3, but changed to a diverging bar chart. Doing so makes the difference between zero and null clearer, this was missing in the previous chart.
It’s just a bit too blocky and for some reason this doesn’t look good to me. It’s so similar to version 3, but I think it just looks worse.
So what did I choose?
In the end I went with a hybrid of V2 & V4.
Version 2 looks the most attractive to me, but you lose a lot of the detail with the overlapping marks. This is fixed by leaving gaps between the bars, and makes it much easier to determine where the years are. Version 4 didn’t look as good, but adding curves to the diverging bar chart softens the look of the chart, without affecting the legibility too much. If the data had gone much higher than five, I probably wouldn’t have chosen curve the ends. One really nice result of this chart type is that years with zero wins appear as a circular dot, almost like a little zero.
If you’re ever stuck on how to visualise a data set, just start building charts and put them onto a dashboard. Once you’ve got something in front of you, you can start comparing them and thinking about what works, and what doesn’t. Sometimes you’ll end up with something completely different to how you pictured it in your head.
This is a great recipe that I’ve used loads of times and is originally from BBC Good Food, I’ve repeated it below as sometimes these recipes disappear from the internet.
It’s a one pot meal which saves on washing up, that’s always a bonus for me!
Main Dish Ingredients
Lamb Shoulder 1kg - 4cm diced cubes
Onions x2 - sliced
Oregano 1 tbsp - chopped
Ground cinnamon, 1/2 tsp
Cinnamon sticks x2 - broken in half
Olive Oil, 2 tbsp
Chopped tomatoes - 400g can
Lamb Stock - 1.2 litres
Orzo - 400g
Parmesan - freshly grated
Rocket - torn
Cucumber - diced
Feta - loads
Red Onion x1 - thinly sliced
Preheat your oven to 180C/160C fan/gas mark 4 and do all the above prep.
In a big casserole dish mix the diced Lamb, sliced onions, oregano, cinnamon sticks, ground cinnamon, and olive oil.
Spread the mixture over the base and bake for 45 minutes, stir halfway through.
Now, pour over the chopped tomatoes and stock, season with salt and pepper.
Cover tightly with a lid (or some foil if you dish hasn’t got a lid) and return to the oven for 1.5 hours, or until the lamb is nice and tender.
Remove the cinnamon sticks and stir in the orzo.
Cover tightly and put back in the oven for 20 minutes, stir halfway.
The orzo should be cooked and the sauce should be thicker. Sometimes, this may take longer than 20 minutes, add some more stock if it’s looking a bit dry and put it back in the oven. (This has taken up to 40 minutes once, but don’t despair, the orzo will cook!)
Serve with a salad
(At times like this, I really wish I’d remembered to take a photo last time I cooked this dish)