I lied. I will, briefly, delve into a small piece of analysis of the Ironman Florida data, something simple and easily comparable with some of the other race data I collected. My intentions were honest, but I’ve been led astray by a reader, Rob Knell, who sent me some graphs – simple frequency distributions showing how many athletes finished within given time frames at both Florida and Hawaii. They are, as Rob put it, an obvious comparison and also quite easy to produce. The charts demonstrate the impact of a selection criteria – Kona Qualification – on the quality of a race field.
Let’s start with Ironman Florida because the data is to hand and we’ll assume the field – age group at least – is typical of an Ironman in general. Three thousand athletes started the race, the charts above show how they performed over the subsequent stages of the race. There is clearly a wide range of abilities in the race, each distribution tapering off to either side of a roughly central peak a long, thick tail out to the slower times. We can see the most common standards in swim, bike and run and note that where both swim and bike have a more discernible peak, run performance is spread broad, perhaps an indication of how important the run is to the final outcome. Swim and bike more tightly packed – possibly the influence of drafting and bunches, at least in part – but on the run the field really breaks up.
The Ironman World Championship is a different beast, the majority of athletes have qualified through performance at another event and the charts reflect this with far narrower, skewed distributions. The focus shifts left towards faster times at every stage with a thinner tail rapidly receding to the final finishers. Performances are, on average, better most noticeably on the run, athletes at Kona know how to pace. If Florida shows us a typical Ironman field then, as we should expect, Hawaii shows us a selection of the best age groupers and pros.
But I’ll add one more race to the list: Challenge Roth. Again I had the data readily available and I thought this might offer an interesting halfway house; a race without selection, but one that attracts fast athletes. The pattern we see in the charts above is not quite as broad as Florida, but neither is it as skewed as Hawaii. It reflects our expectations of Roth, that times are often fast, but there remains a broad range of athletes racing.
There is obviously the potential to debate the impact of differing course and their relative difficulties (or lengths). Or the impact of the size of the field, Florida hosting an additional 1000 athletes above either Roth or Hawaii, are they adding the breadth to its charts? It would be hard to truly judge quality of field from these charts alone, although I’d be reasonably confident citing selection as the cause of that extreme skew in Kona.
Occasionally there is a frustrating moment towards the end of a blog when I realise either my charts or my commentary is flawed. Rob picked me up on one such point, part laziness on my account and part limitations in my statistics, he’s since offered to apply some analysis to the Kona data himself. I adapted his graphs comparing Florida and Kona, using smaller buckets for finisher time to – hopefully – better reflect groupings of athletic ability. I wonder now whether better comparisons could be made using percentages of field rather than absolute frequency or perhaps by normalising finish times relative to some of the fastest athletes in an attempt to minimise the impact of different courses. As often happens with these last minute frustrations I don’t have the time, perhaps I’ll revisit at a later date.
This simple comparison may not highlight anything remarkable or surprise us with its findings, but it demonstrates to me the potential benefits of more accessible race data. It might not change how we train or how I coach, but insights into how races tend to unfold could potentially shape strategies to tackle these problems in training. Of course it would take more data and would need to go deeper than this to begin to address that point. I really should work on my tools for extracting race results.