6

How do you compare swimming and running speeds? For example, is swimming 100m in 60s comparable to running 1 km in four minutes in terms of effort or in terms the times in races (e.g. map percentiles of finish times).

FredrikD
  • 2,181
  • 3
  • 22
  • 42
  • 1
    This is like comparing apples and oranges for me; the only thing you can compare is calories burned. – Baarn Oct 06 '12 at 16:26
  • 2
    This is related to this question. – R. Chung Oct 06 '12 at 17:38
  • Or slightly easier to measure, try to swim at the same heart rate you run at. – Ivo Flipse Oct 07 '12 at 09:10
  • I don't understand the difficulty in comparing speeds? Is it difficult to measure swimming speed? Are you trying to account for swimming upstream vs downstream? In a still body of water, you should be able to simply use the distance travelled / time to calculate the speed just like you would for a runner. – powerpack Oct 08 '12 at 22:12
  • 1
    If you're talking about comparing times, I've found that running is roughly 4x faster than swimming. That is, a top runner runs 400m in ~50s while a top swimmer swims 100m in ~50s. In the same amount of time, the runner covers 4x the distance. This is useful for comparing running times if you're a swimmer or swimming times if you're a runner. – Hartley Brody Apr 28 '13 at 21:10

2 Answers2

10

A similar question was asked here.

That question included data from the 2009 Ironman Triathlon Championship in Kona, Hawaii. As you may know, the race consists of a 2.4-mile open water swim, a 112-mile bicycle ride, and a 26.2-mile marathon. A version of the scatterplot matrix from that question is shown here.

2009 Kona Ironman swim, bike, and run times

The scatterplot matrix above plots the swim, bike, run, and overall times (in hours) for each finisher. Although not shown on the plot, the correlation between swim and run times is .588, which indicates that there is a moderately strong positive relationship between swim and run performance across individuals (though not as strong as the correlation between bike and run times). The plot shows that "good" runners can be mediocre swimmers, and vice versa.

However, if one is interested in the question "what are equivalent run and swim paces for those who both swim and run?" one way to examine the data is to ask "how does the median run pace compare to the median swim pace?" Similarly, you could ask "how does the 90th percentile run pace compare to the 90th percentile swim pace?", how does the 75th percentile run pace compare to the 75th percentile swim pace?" and so on, comparing each percentile to each percentile. This is what the plot below does. It shows a quantile-quantile comparison of run and swim paces, normalized to minutes per mile for running and minutes per 100 yards for swimming. The dotted red lines show the 2nd, 10th, 25th, median, 75th, 90th, and 98th percentile times for run and swim pace.

enter image description here

Note the linearity of the relationship between run and swim pacing. In fact, between about the 2nd percentile and the 98th percentile, the correlation coefficient between the two is above .99. The solid red line shows the regression of swim pace on run pace, and the equation for the regression line is given. As a reminder, this shows the quantile-quantile pacing for multisport athletes who are trying to complete an Ironman, so the pacing will be slower than for single-sport athletes in standalone events.

Community
  • 1
R. Chung
  • 558
  • 2
  • 9
  • great, two questions 1) what is the metric formula? 2) how does the Ironman race compare to the results from Ryan? – FredrikD Oct 08 '12 at 19:10
  • 2
  • Well, if you were interested in, say 10k pace, you could convert from min/10k to min/mile, then use the regression line to get the swim pace, then convert that to, say, pace per 100 meters. 2) This is for multisport athletes and covers a wider range of abilities, so you might think of it as the "equivalent running and swimming pace for someone who does both sports." Ryan was looking at comparing record times for single-sport athletes. And 3) you'll find a slightly different relationship if you had used 2010 data, or a shorter race. I was just showing how one might address the problem.
    • – R. Chung Oct 11 '12 at 18:16
  • This is the best conversion I have seen, but many triathletes are swimmers as an afterthought so comparing a dedicated swimmer with a dedicated runner might bring a different result. Also, a 100m comparison, with water resistance at speed, won't compare closely to distance. – Peter DeWeese Oct 14 '12 at 22:31
  • @PeterDeWeese Agreed, and even doing this for a sprint triathlon might give different results than for an Ironman. Note also that if we looked at the run-swim times rather than the quantiles, the correlation is .588, which indicates that swim time explains about 35% of the variance in run time. But this analysis still gives a way to think about how to answer a question like this. – R. Chung Oct 15 '12 at 02:05
  • @R.Chung Great answer and approach! Did you do a weighted regression or a simple linear regression (SLR). For run vs swim it looks like the homoscedastic assumption of an SLR was violated, this will give too much statistical weight to slower individuals. The individuals who were espeically slow swimmers likely have too much leverage on the regression line. – Rider_X Jan 07 '15 at 22:56
  • @Rider_X I don't remember exactly but I suspect that I did OLS regression. As for heteroscedasticity, recall that OLS produces unbiased though inefficient estimates so the coefficients will be right on average but statistical inference may be off. In any event, I was just doing this to show how one might approach the problem -- you might want to also look at this related answer. – R. Chung Jan 07 '15 at 23:39