To extend on our conversation on your blog site, here are some thoughts on 400m hurdles from a lecture I delivered in March 2011. Through conversations with Wynford Leyshon and data extracted from Dai Greene’s dissertation, I compiled a workflow to extract the key contact times from within the race. I conducted the exercise on a sample of 12 performances in mid to late 2010.
I extended the traditional split points of the first touch beyond each hurdle to incorporate contact points of the two strides either side of each hurdle. This was thought to generate not only an indication of flight time, but also multiple derivatives of timings around the barriers to explore.
This data was compared to a model of performance for split times and also against normative performance data accessed for the athlete concerned. Crudely, I took a mean of 5 competitive races and calculated a ±0.05 band to be sensitive enough to provide an indication of performance for the section time hurdle (+1 to +1). This value was used to provide the tricolour conditional format feedback in the dashboard.
The absolute difference between -1 and +1 contact points were used to calculate flight time. On reflection, this should have been expressed to 2 decimal places, as the data collection procedure is not sensitive enough to deliver this accuracy.
The section times were calculated as the differences between the +1 contact points, and presented in the same fashion as your graph (but as an area graph). The same data was presented in two ways in this dashboard; I felt they were both important visual representations (conditional format and area graph). The dual axes plot also showed the flight times as they were potentially linked to the section time. The sample illustrated a couple of interesting features, such patterns of flight time differences with alternate lead leg take-off in the example below (blue line zigzagging up to H5).
The comparisons with the modelled performance helped to show relative pacing strategies adopted by different athletes in different races. This example shows an athlete who was running quicker than the time he achieved throughout the race, but could not maintain this speed endurance and fatigued at the latter stages.
When comparing the same athlete over a series of races, it highlighted there is not necessarily a consistent pacing strategy adopted in different races. Based on the modelled performance to run a 48.15s race, the line graph show three derivatives of racing strategy. Red = slower starting pace which did not accelerate to convert to a quick time, Yellow = quicker starting pace which was not maintained, Green = controlled starting pace with an increase in pace after H7.
This methodology requires more rigorous exploration as it could be addressed a multiple levels. It may require scrutiny of the modelled parameters applied in this instance. Or is the scale of difference too small to even warrant exploration in the first place? I am looking forward to supervising an Undergraduate dissertation in this area in September to explore it further. Hopefully, this student will be able to share his thoughts and findings this time next year?
I am an advocate of the dashboard summaries of performance and offer the final slide from the lecture to close the blog… the overall dashboard can be seen here.
Mike: I hope this was a useful exercise for you to undertake…