The Consistency of Consistency

In biathlon, consistency is something most athletes are looking for, ideally from one season to the next, assuming the performance in a certain metric is at the level they are happy with. I built a dashboard in Tableau Public that looks at the career and seasonal form, averages and variance, and at consistency for the following metrics:

• Prone Shooting
• Standing Shooting
• Total (combined) Shooting
• Ski speed (in Km/H)
• Ski Score (Z)
• Rank
• Shooting Time Score (Z)
• Range Time Score (Z)

From the RealBiatlon.com website: Z-score (Standard score) Number of standard deviations by which metrics are above or below the mean (based on back from median data)

The data used goes back to the 2016-2017 season, so when I refer to career averages the data will not include any data from before the 2016-17 season. To highlight this I have used an asterisk whenever using career. Please note that when using different metrics like this, the meaning of above zero and below zero is not always positive or negative. I.e. Z scores for skiing are better when negative (meaning below average) but for shooting percentage the higher number the better.

As examples often are a good way of explaining visualizations I am going to start with Lisa Hauser, and her Ski Score (Z).

This simply shows Lisa’s average for Ski Score (Z) and the sharp drop for the current season clearly stands out, meaning she went from a just below average skier to a faster than average skier. Also, we can see she has been much faster than her career* average, indicating she must have really focussed on her skiing the last preparation. Has that affected her shooting? Let’s see by changing the metric to Total (combined) Shooting and look at…

This tells us that her current season’s average and her career* average are almost identical, so no change here. We can also see that as the season progresses she is seeing better results (for shooting percentage, higher is better).

Now can we get more out of this? The following shows the difference between actual results and the career* average and shows it cumulatively, based on the assumption the multiple bad results in a row, even with a good result between a number of bad ones, has a bad impact on form.

Due to her less than ideal first number of races (with regards to total shooting) and a lesser performance in the last race of the previous season, the chart shows a lower than desired profile, that however sings upward towards the current status of the current season.

One could argue however, that the seasons are separate entities, and the end of last season would not impact the form of an athlete at the start of the current season.

The same applies in this case for the current season, showing the bad start and the incline due to better results in the second trimester, but the previous season now has no impact at all. A better example of showing a differnece between career* and season is the follwing for Shooting Time Score (Z):

If we want to see more about consistency, the metrics are used in absolute form. It doesn’t matter if a result is good or bad, as long as it differs from the previous results it introduces inconsistency. So the next chart shows the absolute values of the differences between actual race resultes and season averages.

Now the hight (or depth) of the chart shows the size of inconsistency, where the direction and steepness show how much the race result impacted the consistency.

Lastly to satisfy the more statical inclined readers below are the Variance charts, showing the spread of results and the average Variance per season (still Lias Hauser’s Shooting time score (Z)).

This dashboard is not coming to a specific conclusion, but rather a tool to further research an athletes’ performances, form, and consistency, intended to be used interactively by you! So go have a look and have fun with it.