[Note: This is one of a series of posts on the
2013 CrossFit Games Open. For more details on the project, check out this post here. I previously wrote on height and CrossFit here.]
I’ve arrived at many a “functional fitness” (code for not trademarked Crossfit) competition to discover that all the other entrants are in their early twenties. What’s an old guy to do?
As part of this look into the results of the 2013 Crossfit Games Open, I wanted to examine the effect of age on athletic performance. What happens as we age? When does performance peak? Did any one event favor older competitors or vice-versa?
- Men’s scores on average peak at age 23. Women’s scores on average peak at 25 years old.
- Women’s performance decreases until they reach 30, at which point the rate of decrease is smaller until their mid-30s.
- Men lose on average 3% of their ranking in the open each year from age 22 to 39. Women lose 1.8% of their ranking on average from 25 to 39.
- Older competitors of both sexes performed better at 13.5 (never-ending Fran) compared to younger participants. For men, 13.2 (Box Jumps+Deadlifts+Push Press) showed a skew towards better performance for younger male competitors. For women, 13.1 (Burpees + Snatches) skewed better for younger female competitors.
*The Data: Age & Crossfit *
Looking across the entire Open, we have the following summary of ages across all men:
Min. 13.00 1st Qu. 26.00 Median 32.48 Mean 3rd 37.00 Qu. Max. 80.00 NA's 1 [
And for all women:
Min. 14.00 1st Qu. 26.00 Median 31.00 Mean 31.99 3rd Qu. 37.00 Max. 79.00 NA's 2
As the entire set of entrants in the Games Open this year was unwieldy (over 200,000 entrants), I broke it down into manageable samples. This is the sample of 10,806 Men chosen at random (all of whom completed all 5 workouts):
Here’s a similar summary for women:
Interesting to compare the two distributions. While both women and mean peak at 28, women drop off much more suddenly compared to men who have a smaller decline as they approach their mid-thirties. Notice all the peak in participation around cultural age milestones such as 40.
So far, so good.
Performance vs. Age
Let’s plot the average rank (lower being better) versus age for our sample of men:
The performance is best, on average, for younger men. In this sample, the average is lowest for men in their early 20s. The inflection point of our average happens right at 23 and then goes up from there as men age.
We can zoom in further to the age range of highest performance (18-35):
Peak performance maxes for the population around 23 and goes down from there. (The gray stripes are because making graphs is hard in R. Sorry!)
Let’s look at the sample of 10,600 women. Here’s their age vs performance:
(Note the dip after 50 is likely due to the weights used changing for women in the super-masters divisions.)
And limiting our age range to find the age with the lowest average score:
Women’s performance is peaking at age 25 on average.
Predicting Performance as a Function of Age
Now that we know where performance peaks (23 for men, 25 for women), can we model out how much of a disadvantage are competitors as they age?
Let’s start with the women this time. Here’s a linear regression for women 25 through 49:
Using this linear model, we can predict the average rank of participants aged 25 would be 14,487. Each successive year would add 265 places (using this model) 265 places. Or, 1.83% per year slower. Thus, a 39 year old on average is 25.62% worse ranked than a 25 year old woman.
For men, we can predict that the average male will be at place 21,249. Each year will add 680 places or about 3 percent until the average 39 year old is finishing 9,739 places (45% higher) beyond the average 23 year old.
Each Games Open Event Versus Age
I next plotted the average ranking for each event in the Open (13.1 -> 13.5) for both men 18-49:
Above we see that 13.2 (box jumps + push press + deadlifts) was particularly punishing on older male competitors while younger competitors performed better on average. At the height of the box jumps for this event (24″), I believe younger legs would likely have made all the difference.
Plotting the same for our sample of 10,800 women:
- As we can see from the performance graphs, the performance as a function of age doesn’t look linear. Sinclair Coeffcicients use the top performances at different weight classes to then generate multipliers. It’s clear that something similar is the correct methodology to generate a good equalizing model for age in Crossfit competitions. Perhaps that’s the topic of the next post.
- Also, getting old appears to suck. 😉