Sunday, February 24, 2013

Energy saving while drafting (IM Focus)

So it's now less than 4 weeks to IM Melbourne in Australia, and just in time for a somewhat light hearted look at drafting, or conserving energy legally, whichever you prefer to be the correct term.

It's been widely accepted that there is benefit drafting behind another rider, in fact Kyle(1) published data as early as 1979 showing a 47% reduction in drag at 0 m behind another cyclist, and 27% reduction at 2 m.

Some time ago our local triathlon forum had two schools of thought emerge on "pacelining" in IM, where to be paceline "legal" a cyclist has to be 12m from the cyclist in front; the first being there was no benefit at 12m, and the second that there was benefit but that it was fairly minimal.  Knowing that there would be some benefit, but unsure how much, I designed an experiment that would try to test the reduction.

This involved mounting a laser pointer on the bike (Cervelo P3) pointing at the ground 12m away from the front of my bike.  When following another rider, all I had to do was make the laser dot on the ground match the cyclist in front's back wheel and bingo accurate measure of the correct drafting distance.  Since this test was done on a velodrome, it was pretty easy to concentrate enough (and safely) to ensure that I was pretty accurate with the trailing distance (perhaps +/- 0.2 m error over the interval durations).

But I was concerned that I would possibly be biased in the way that I extrapolated the results; I believed the benefit to be there, so I might unwittingly find what I was looking for without knowing that I had somehow biased the analysis.  The solution was to get friend and previous coach Alex Simmons to do a blind analysis of my power file WITHOUT telling him what I was doing, only that I'd done a number of intervals and that sometimes I was in front, and sometimes not. Each interval had no intermediate or lap markers to denote changes from in front to drafting, and Alex did not know whether I started in front or behind on any of the intervals.

My quick analysis of the results was enough to confirm what I thought, notably an approximate 10% reduction in power required at 12m separation (front wheel to front wheel), which is pretty darn significant, and could mean the difference between running the IM marathon of your life, and lying huddled on the side of the road somewhere imbibing flat coke in a desperate attempt to get to the finish.... but I digress.

Alex, however, went to town on my file; presented with a challenge he left no scientific stone(2) unturned in the quest for knowledge....


He found (to cut to the chase):

So in summary, the gain by drafting the other rider was a reduction in apparent-CdA of:

Interval 1: 0.035m^2
Interval 2: 0.033m^2
Interval 3: 0.026m^2

In terms of energy benefit for for Rob when drafting over leading, when riding at 40km/h this equates to wattage savings of:

Interval 1: 29W
Interval 2: 27W
Interval 3: 21W

The intervals ranged around 240, 260, 280 watts average in respective intervals 1, 2, and 3.  So a very handy saving - roughly 7.5 - 12%.

How significant is that you ask? Well if you're typical triathlete drag (CdA 0.285) at fairly typical triathlete race watts for IM (190) and you got 10% free speed (209 effective watts) your speed would increase from 9.71 m/s to 10.08 m/s, sending you 0.37m (1 ft 3 inches) further down the road for EVERY SECOND you are on the bike!  Your bike time would drop from a hypothetical 5:08:58 to 4:57:37 (3).  A whopping 11 minutes and 20 seconds saved. 

OK, so after all this nerdy work what would be great is if we could get some practical advice on how to use this to our advantage in an IM cycle race. Assuming you have a power meter, and you know what average watts you intend to target for the entire race, I've scribbled up some cheat sheets give you a pictorial and rule based guide to using the ~10% energy saving you'd get from drafting at 12m as you interact with other riders.  It's available in PDF in both right hand drive (Aus, NZ, UK, Japan etc) and left hand drive (Euro, USA & Can etc) versions.

Have fun out there!

Big thank you to Alex Simmons for taking the time to write up the results with proper analysis and detail.
You can read the full analysis on Alex's blog.



Kyle, C.R. (1979) Reduction of wind resistance and power
output of racing cyclists and runners travelling in
groups. Ergonomics, 22 (4), 387–397.


R.A. Lukes*, S.B. Chin† and S.J. Haake*

The understanding and development of cycling aerodynamics,
ISEA Sports Engineering (2005) 8, 59–74

* Sports Engineering Research Group, Department of Mechanical Engineering, University of Sheffield, UK
† Department of Mechanical Engineering, University of Sheffield, UK

Note: Lukes et. al. is a great read (not overly scientific) and a copy is available on this site.

(2) Science: it works b*tches.

(3) Yes it is a hypothetical calculation, and no you can't hold exactly the same aerodynamics for an entire IM bike leg. Despite this, of the relatively few predictions  for IM based bike times where I've known the rider characteristics (3 total), I've been within a minute or so of the riders actual time.

Saturday, February 16, 2013

Confused aero testing

So inside triathlon have published a new aero test on some big name aero bikes... and come to a conclusion that one particular model is 53, 81 or 153 seconds faster than the competitors over a 70.3 bike leg.

What's interesting and slightly confusing is their statement immediately following the results:

"As stated in the printed article, the test is imperfect, as are all bike aerodynamic tests. Several factors—most notably the influence of a rider—could make the real life performance of these bikes different than the results of Inside Triathlon’s test."

Er, well, yes next time I want to send the bike round the course by itself I'll bear these tests in mind... so why are we testing like this again??  Tunnel time isn't cheap, and sweeping through yaw also complicates things, particularly with drive and non drive side aero that varies between the sides tested.

Real world aero tests where n=1 and n=me obviously do have a lot of benefit and should, of course, not be included in "all bike aerodynamic tests".

Perhaps they're still trying to figure out how to use and interpret aero tests that actually have meaning... kudos for doing that, but there are other ways to design an aero study.

Side notes:

The drag estimations they recorded did have variations in drag at zero yaw after swinging through +20 to -20 and back to 0. A better way of presenting results would have been with the error(s) associated with each measurement.

Trek Speed Concept 9 Series: 435 grams +/- 18 grams
Cervélo P5 Six: 495 grams +/- 10 grams
Specialized S-Works Shiv: 525 grams +/- 28 grams
Orbea Ordu GDi2: 606 grams +/- 19.5 grams

Even better, some statistical analysis of the results including for example, standard deviation, standard error or even better a confidence interval (we're 99% confident the value is within a certain range).  It's unclear how many measurements were actually taken.

Secondly, they've weighted the yaw results *again*.  (Sigh).

"In addition to the amount of wind resistance, yaw angle also changes with rider speed (faster rider, shallower yaw; faster wind, wider yaw), so we calculated the fraction of time a rider would spend in various yaw angle ranges when riding at 23 miles per hour and weighted the drag created at each yaw angle accordingly. These were the results for a rider traveling at 23mph in 8.1mph wind."

Er, does this mean you have a course where the rider spends equal time in every compass direction?  I don't know about you, but often race courses are out and back on a single stretch of road that will spend a large amount of time in certain directions of yaw.  They've collected met data from 49 major cities from Seattle to Miami and arrived at an average of 8.1mph of wind... perhaps they should have also analyzed a variety of courses and determined an average yaw value percentages based on typical conditions for those sites.

Weighting results completely obfuscates the value of testing at yaw.  Objects performing at high yaw, may not perform as well at low yaw and vice versa.  When combined the final results may be closer by the nature of weighted averaging, but may not be representative of real world experience.  It would be better to present cases for zero, low and high yaw situations and allow people to have the data that may influence choice in those conditions... certainly you might not swap a bike under low or high yaw situations if you only have one, but other choices like helmets or wheels might easily be changed given different conditions.

"Ideally, an athlete would measure real wind speed on each individual racecourse and run a calculation to find the drag they expect to face during a race then pick accordingly from their stockpile of race wheel. This is, of course, not practical for most athletes so we use this approximation."