While ten months may seem like a long time, it isn’t long for elite swimmers whose preparations for the 2020 Tokyo Olympics are already underway. Though their season has just begun, many top collegiate swimmers have their eyes set on the 2020 Tokyo Olympic games. For some of the almost sure-fire Olympians – such as Stanford and Canada’s Taylor Ruck – that means taking a redshirt year off of NCAA competition to fully concentrate on long course training. For other top NCAA swimmers, that might mean concentrating winter training on Olympic events and long-course training. And for many it will mean more focused training and fewer vacations over the 2019 summer.
With a couple of different training changes in play, what the overall effects of Olympic-year training will mean for NCAA competition is unclear. Swimmers redshirting and taking the year off to focus solely on Olympic training would certainly mean less high-end NCAA talent. Swimmers concentrating more on long-course training and events might also mean that short-course times will be a bit slower. On the other hand, Olympic years tend to convince more swimmers to train harder for longer – maybe taking shorter summer vacations or other breaks from swimming.
I examined data from the top NCAA swimmers from the last ten years to see if training during Olympic years had any effects on their times. As I’ll show, while the overall effect isn’t significantly faster or slower, there are some trends in the top swimmer’s times.
The above chart shows the average of top 16 times across all events in each NCAA season since 2010. USA Swimming has data going back to 2008 on NCAA times, but 2008 and 2009 were excluded due to the differences in suit technology. The top 16 times in each event were compared to 2009’s top times as a benchmark. A value of 1 for a given season means that the top 16 times were on average the same speed as in 2009 and a value below 1 means that the times were faster than 2009. The 2012 and 2016 are Olympic seasons and are bolded for emphasis.
As you can see, there is a clear yearly linear improvement in both men’s and women’s top 16 NCAA times. The 2012 women’s season – an Olympic year – is clearly faster than expected per the yearly trend. However, the other Olympic seasons look more in line with non-Olympic seasons. To try and measure if the Olympic seasons overall were in fact outliers, I plotted the same data without the two Olympic seasons in 2012 and 2016 and re-applied a linear fit to the data.
The overall trend of times is similar, but the data is more linear when the Olympic years are excluded. This indicates that the Olympic years are outliers. To drive this point home, the following chart compares the R squared values – a measure of how linear the data is – in all seasons versus seasons excluding the Olympic years (2012 and 2016). A value closer to 1 indicates the data is more linear. As can be seen below, the men’s times were relatively linear with or without the Olympic seasons whereas the women’s times were significantly more linear without the Olympic years, meaning that the Olympics overall had an effect on Women’s NCAA times.
One possible reason that Women’s NCAA times are more affected during Olympic seasons is that Women are more likely to be Olympians during their time in college than men are. The average age of female finalists at the last few Swimming World Championships was around 22 and male was around 23.5 years old. It is reasonable that women’s NCAA times would be more strongly affected by Olympic preparation than men’s since more potential female Olympians would be competing in NCAA meets.
Lastly, I looked at different segments of women’s swimmers to see if the Olympic-effect was stronger for faster swimmers. Shown below is a similar measurement if Olympic seasons were in general outliers – this time for 1st place, 16th place, and 100th place times averaged across events. As you can see, the faster swimmers – the 1st and 16th place times – were more affected by Olympic seasons training whereas the sub-elite level swimmers – represented by the 100th place time – was not significantly affected. Which makes sense since swimmers more likely to qualify would be more likely to heavily alter their training patterns to give themselves a better shot at making the Olympic rosters and medaling.
What this analysis doesn’t show is if the times were overall faster or slower. The data didn’t show a clear overall effect in terms of the times being faster or slower, just that they were outliers when compared to non-Olympic seasons. Thought the 2012 women’s NCAA season was clearly much faster. While I don’t have data for it here, I believe that the Olympic effect on times is trending less and less so as the average age of Olympic contestants has gone up with more opportunities to continue swimming professionally after college. But the effect is still clear and something to watch out as we get into the 2020 NCAA season.
Sorry, can’t help myself
* R2 is not a test of linearity, but of how well the Xs predict Y. Add in higher-order polynomial’s of X, and R2 will go up.
* Are the Olympic year “outliers” really outliers? Why not test if the 2012/2016 years have a statistically significant deviation from the trend?
* Aren’t the negative slope coefficients in F1/F2 clear evidence that times are getting faster?
In a first degree polynomial such as the one Kevin has used, ‘how well the Xs predict Y’ is exactly a measure of linearity.
R2 is a measure of how accurate an estimation model is. In this case the model is a linear – so more linear would mean a R2 value closer to 1.
The effect on an R2 value is the easiest way to determine outlier status of a few correlated data points on otherwise linear data. One could also do a T-test of the outlier olympic years against other seasons. But that test would only be meaningful after accounting for the yearly trend towards faster times and the math is much more difficult.
Right, NCAA finals times are clearly getting faster post suit years.
It would be interesting to see if there is statistical significance to the “step down” from 2013-2014 and 2016-2017. If so, what are the origins?
It’s almost more interesting to me to note the overall trend of roughly a quarter of a percent improvement per year relative to 2009 times. Couple of questions that I’m just putting out there as hypotheticals, so please don’t feel pressured to put in any more work if you don’t want to:
Is this improvement more or less consistent across all events or have we seen more relative improvement in the 200s vs. the 100s etc.?
Is this improvement consistent when looking at the top 16 of long course national performances? The age distribution of meet qualifiers is different, and the pro-swimmer time trajectory is often different in non-Olympic years, so I’d be surprised if it was very close, but… Read more »
Not sure about the 100 vs. 200 question since I wasn’t focusing on the improvement aspect, but clearly something interesting to look at in the future. I have always found it curious that every single SCY NCAA record has been broken since the suit years (except for Auburn’s 200 Free relay), but only about half of the LCM records have fallen since then.