Tableau For Sport – Identifying Signs Of Fatigue
Sports clubs spend millions of pounds every year on the latest technology and innovations to aid the process of fatigue management and injury prevention. With each new technology comes more data, more metrics and often more confusion. Two things that my experience and my learnings from others has taught me is this:
- Concentrate on the metrics in which you have a strong understanding of their impact
- Athletes need to be treated as individuals, with individualised plans and ‘person-unique’ targets
With these two in mind, let’s take a look at how we can utilise Tableau to give us a quick understanding of an individual’s status or performance that will provide us with actionable insight. This in turn can enable us to make alterations to an athlete’s prehab, training and lifestyle to reduce the risk of injury.
Let’s Take A Look At The Data
As I always emphasise, the quality and structure of the data is vital and if (like many sports clubs) your datasets are in Excel files then please read this article first: ‘Preparing Excel files for analysis’.
So here is the data I am going to be working with in this example – fictional hydration data using some of the best football players of years gone by:
Connecting To My Data
Firstly we need to connect to our data, so I am going to use the ‘Add New Data Source’ button on the toolbar and I am going to select my hydration data.
Viz My Data
First port of call is to create the base of our viz, and I am going to follow my rule number 1 when it comes to human performance data by using a fairly well understood metric – hydration.
I have dropped the ‘Date’ field on to the columns shelf and ‘SUM(Hydration Score)’ on to the rows shelf. I have also added the ‘Player’ field to the filters shelf and changed the filter to a single value slider, so we can look at one individual player at a time.
Now it’s time to add a bit of context and some insight….
To do this I am going to follow my 2nd rule mentioned earlier and, instead of putting in a generic target line or team average, I am going to consider each player’s individual scores and their variation from their norm.
There is, like most things in Tableau, many ways in which to achieve this. On this occasion I am going to make use of a feature that was introduced in Tableau 9 – the analytics pane.
By selecting this pane we are provided we some very quick and intuitive analytics objects, including reference lines, trend lines and forecasting to name a few.
What we are after from this analytics pane is a distribution band. So I am simply going to drag the ‘Distribution Band’ object from the analytics pane on to our view and as you start to drag you will see a little box appear which enables you to specifically choose the scope of the distribution band. In this example it doesn’t matter which we choose as we will get the same result with any of the choices. To understand more about distribution bands and references lines then click here.
From here we will be provided with a selection of options regarding both the type of distribution and the formatting of the band.
So for this example I wanted to to use the standard deviation and the first band I want to create is between -2 and 2 standard deviations from the norm. This is also the point at which I can change the label, the fill colour and line colour.
Once I have done this for everything within 2 standard deviations for the norm, I want to replicate this band but for everything within 1 standard deviation. I shall do this by following the exact same steps as above but this time in the ‘Factors’ dialog box I shall type “-1,1”.
At this point it is important to note that I created the 2 standard deviations band first as each new band added will be placed on top of previous bands, therefore if we applied the bands in the opposite order the 2 SD band would hide the 1 SD band.
So we now have a chart with distribution bands for everything within 1 and 2 standard deviations of that players norm. But one key thing we are missing is to emphasise all marks (hydration scores) that are more than 2 standard deviations from the norm.
We could do this by adding in another distribution band, but what factors would we use – 3 standard deviations? 4 standard deviations? What if there is the chance that there could be massive fluctuations in your data?
To get around all of this I avoid using another band and instead I right click on the chart and select ‘Format’. From the formatting pane I have then changed the chart pane’s shading to the required colour.
And there we have it, a simple line chart of hydration scores over time, but by adding in our distribution bands we can quickly and easily identify when an individual has fluctuated significantly from their norm.
This can then enable us to take immediate action and reduce the risk of injury through intervention.
Take a look at the below chart as an example, this shows the (fake) hydration scores for Zinedine Zidane and we can clearly see that the most recent scores show a large variation from his norm, potentially placing the player at a greater risk of injury and enabling the human performance / sport science team to intervene after delving deeper and carrying out further analysis. At this point I am keen to stress that this is only one indicator of a potential increase in risk of injury and should be taken in to consideration alongside other metrics.
I have used a grayscale shading scheme in order to not alienate the colour-blind audience. I know that the general consensus within many sporting environments would be to use a red-amber-green colour scheme, so I have recreated the same chart with this traffic light shading below.
