Of all the major1 sports, cycling “performance” is the most easily quantifiable. There are power meters fixed to the majority of the bikes in the professional peloton and wind tunnel testing has become a staple of the offseason regime for folks with pro contracts and those with cash left over from their latest Audi purchase. Yet (and I promise I’ll get off my high horse after this), people forget that “performance” is not about the Watts, the deep rims, or the obnoxious helmets; it’s about the results. (which is fortunate, because power data for many pros is simply a matter of speculation2).
This brings me to my first assumption: a rider’s performance is directly related to their position on the results sheet. A cyclist needs to have the power, the handling skills, and the tactical awareness to get high up in the results (not to mention a little luck). The limitation this assumption introduces is the fact that the work of the lead out men3 and the domestiques go unaccounted. I’m okay with this assumption because domestiques and lead out men are typically not nearly as good as the person they’re leading out, so would generally fall to less significant places if riding for themselves.
Let’s start with the 2013 World Tour. Up until the end of October, 228 riders have scored at least one World Tour point. Their ages4 fill the distribution in Figure 1.
We observe a roughly Gaussian distribution, peaking at 28 years old. However, this doesn’t account for the relative performances of each age group (i.e. a rider with 1 WorldTour point and one with 100 points count the same, but shouldn’t). To give a better representation of the performances, I took the sum of the points scored by each age group. The results are shown in Figure 2.
I think this distribution gives a much better approximation of how performance is related to age because it accounts for the magnitude of results, not just the count (though a count is indicative of the trend by itself). That being said, I think Figure 1 is essential for a qualitative interpretation of Figure 2, as the total number of World Tour points scored and the number of riders scoring those points are obviously correlated.
Because these data are quite noisy (we would expect nothing less from a sample of just one year), I repeated this process for 2012 and 2011, the first year the current system we implemented. These results are shown in Figures 3 and 4.
In both figures, we observe that a Gaussian distribution better fits the data, though for the latter figure, there is significant deviation at age 32. This spike has a lot to do with the likes of Bradley Wiggins, Tom Boonen, Alejandro Valverde, and Simon Gerrans having banner years in 2012, taking 4 of the top 6 positions, not to mention Joaquim Rodriguez and Michele Scarponi’s great 2011 seasons putting them 3rd and 4th, respectively, in the WorldTour for that year.
However, a lot of the noise in the distribution may be a result of the way the World Tour ranking system is designed. The “worst” cyclists on this list (and there are quite a few), have only one point. Compare this to Philippe Gilbert in 2011 where he accumulated 718 points and enough wins to keep even the most ambitious amateur sandbagger content for their whole “career.” Does this make him 718 times better than those with only one point? I think not.
I turned to Pro Cycling Stats for a ranking system that put the performances of different professional cyclists in better context.5 They have their own points per age distribution that I encourage you to check out,6 but I feel that once we start getting past cyclist #300, the ranking starts to get a bit muddied as it’s so terribly hard to properly weight races and account for riders across country lines. Therefore, I stuck with a similar sample size as the World Tour – 250 riders. In Figures 5 and 6, I add the data from Pro Cycling Stats to Figures 3 and 4, normalizing both point values in the latter figure.
In Figure 5, we see that the Pro Cycling Stats and World Tour populations exhibit a nearly-identical trend. However, in Figure 6, we observe that the Pro Cycling Stats model has significantly less noise, due in large part to the nature of their points-allocation style.
If we imagine a Gaussian curve over the data, we might conclude from the data that peak cycling performance is achieved between the ages of 26 and 28. That being said, I think it’s quite likely 2011-2013 were just bad years on the whole for riders aged 29-31, and that as the years progress and more data is presented, we might see performances of the middle-aged (relatively speaking) cyclists approach or even match those of their younger competitors.
Pro Cycling Stats has ranking data as far back as 2005, but as far as I’m concerned, every performance in those years should be subject to significant skepticism. It’s like comparing football quarterbacks from different eras – there’s just a completely altered climate that inserts so much murkiness (e.g. older riders might have been better at doping and getting away with it).
As expected, there is a clear correlation between cyclist age and their performance as measured by results-based ranking systems. From the last three years, the data indicates that the average rider will reach their peak around age 27, though it will be interesting to see how the distribution fills out as the seasons progress.
As I was compiling the age information for many of the cyclists, I noticed that the birthdays tended to be earlier on in the year. If so, perhaps this is a byproduct of the ‘racing age’ organization that cycling has adopted. If you’ve read Outliers by Malcolm Gladwell, you will be familiar with this phenomenon. Athletes born earlier in the year are more developed than their slightly younger counterparts, and therefore get more attention and support from coaches, amplifying this difference all the way to the professional level. Perhaps a subject for another analysis.
Thanks for reading.
1 I know, I know, cycling hardly qualifies as ‘major’ in the States, but from an international perspective, I have no qualms calling it ‘major.’
2 If I never heard or saw the term ‘VAM’ again in my life, I think I would die a very happy man.
3 And women. This analysis focuses on the male professional peloton as there are more (and better recorded) data.
4 ‘Age,’ used throughout this post, means racing age, the oldest age one will be in the current year.
5 You can find all the rankings here.
6 You can find Pro Cycling Stat’s age distribution here.