There is a commonly held belief – and not one I disagree with – that age group Ironman performances, most specifically front-of-pack age group Ironman performances, are getting faster every year. Much as I share this view point I was also interested to see if the Ironman results reflected it – whether the perception that races are getting faster is actually visible in the available data. Over the last couple of months I’ve occasionally played with that data, testing ideas and seeing if I can determine a trend. I decided to put these ideas to use on fresh Ironman Florida results. My conclusion: some age groups are getting faster, but not all and not consistently or uniformly and not necessarily significantly.
To keep things simple I’ve focussed on the question of speed, not whether the field is more competitive or Kona qualification harder. They relate and I would expect the answer to both to be much the same, but their causes and the analysis required are slightly different. Speed is simpler. If the age group field is getting faster then we’d expect finish times now to be faster than finish times 10 years ago. It’s not quite that simple of course because race times reflect not only the quality of the field, but also the difficulty of the race; course changes and conditions contribute to the overall outcome on race day. I’m only looking for the presence of a change, I’m not concerned with attempting to quantify it (that sounds far too difficult).
Divide results up by year and position in an age group and we can fit a trend line, a linear regression, to those points (as in the graph above). What we might then observe is that for these plotted age group positions there is often, but not always, a negative slope to their regression lines. That matches the casual observation of age groupers getting faster – at least at the front of the pack – but it should also be noted that these lines are poor fits to the data and the trends are hardly significant.
Had my date range for Ironman Florida been 2004-2011 then the times for the top 10 M40-44 age groupers would show a slight positive trend and were we willing to overlook the poor fit we might happily conclude the age group was getting slower. The graphs and regression may reflect why the observation is made, but the data isn’t sufficient to state whether this is a genuine shift in performance or chance variation in race conditions and the age group field. I continue fully aware that we are unable to make that distinction.
We may also note that not all age group positions show a negative slope, 50th place in the M35-39 age group shows the opposite, a small positive slope. Not every trend is faster. I can plot the slopes (as in the graph above) from each regression to see how they vary through the age group and also by how much. Sure enough, outside the top 20, none of the M35-39 age groupers show a trend to faster times and even within the top 20 many of those trends are small. If we look back at the regression line for 10th or 20th place in the M35-39 age group the magnitude of the slope is insignificant. On the other hand in the M40-44 division while the trends may not always be particularly steep they are all consistently towards faster times.
I like approaches that enable me to quickly visualise the differences between a wide range of age groups. A histogram using the same age group position and respective regression data can offer that. In the charts below the histograms indicate the distribution of positive (slower) and negative (faster) slope values within a given age group; yellow including the whole age group, red just the top 20 positions (for some categories this is every athlete). Note that the x axis is inverted to place faster trending, negative, values on the right.
Here we see that at least for Ironman Florida there is a divide. Some age groups showing mostly trends towards faster times, some showing mostly trends towards slower times. For the men 40 seems to mark a shift from trending slower to trending faster. Women appear to make the change sooner with 35 apparently marking this shift. In most cases the majority of slopes fall close to 0 (horizontal) indicating that the slope is shallow. If these changes in age group performance are real they are also generally quite small over the time frame. This is also true for the top 20 athletes.
I didn’t leave it there. This post is growing long, but I have just one more thing.
This approach can be performed with any race for which I have a reasonable number of years of results (I chose 5 as a reasonable number) so of course I did (click for larger images).
Curious about that 40 year old divide I plotted values from the M30-34 and the M40-44 age groups to see how they compare. There are mixed results: in the M30-34 age category some races show most trending towards faster times, some towards slower, whereas for the M40-44 division there is much more weighting on the faster trend side.
Always bearing in mind that the presence of a slope on a poorly fitting linear regression is not a great indicator of significance we may still suspect that 40-something age groupers are getting faster. We might also suspect that this isn’t the case for every age group, or for every type of athlete, fast or slow, within an age group. Along with that we may even note that the trends hinted at are often relatively small.
I can’t claim to have provided proof. There may be (are almost certainly) better ways to analyse this. However I am happy to continue believing that front-of-pack age groupers are generally getting fast; which with diminishing numbers of Kona slots at individual races means the competition for a slot is also likely increasing.