The pursuit of the sub-2 marathon: Where to next?

08 May 2017 Posted by

Yesterday, I started writing what was going to be a shorter piece for a magazine or newspaper here in SA, but as sometimes happens, it turned into a long one, so I’m publishing it here instead.  Apologies for length, but it allows me to get into some technical details based on some nice Twitter conversations last weekend, and to use this as my final word on the Sub-2 marathon attempt and Kipchoge’s performance from the weekend.

In this piece, I ask where we go next, whether “human performance limits have been recalibrated”, and whether we can expect things to speed up or slow down?



The summit in sight, where do we go next?

If the sub-2 hour marathon is the Mount Everest of human endurance performance in 2017, the Eliud Kipchoge has reached the ledge just beneath its summit.

Kenya’s Olympic marathon champion, and the world’s number one marathoner, came within 25 seconds of dipping under the two-hour mark in the staged Nike test last weekend.  That equates to a shade under 150m short of running the distance in 1:59:xx.

Expectedly, the performance has inspired much excitement.  One report called it a “recalibration of what is humanly possible”, and the already lively debate about the prospects of that sub-2 will only intensify as a result.

But where do we go from here?  What happens next in the quest to summit this mountain?  To continue the analogy, now that Kipchoge has shown the way and put the ropes in place, how long will it take before another runner, or Kipchoge again, returns to finish the climb and reach the summit?


Two perspectives – sub-2 or a slowing down?

That depends on how flexible you want to be with your standards for what constitutes a legitimate marathon performance.  The Nike event was highly stage-managed, and a few of the tactics used to “optimize” the conditions for Kipchoge make it ineligible as an official world record.  This is why the world record still belongs to Kenya’s Dennis Kimetto, who ran 2:02:57 in Berlin in 2014, even though Kipchoge is now 2:33 faster over the distance!

The size of that gap between Kipchoge’s “theoretically optimized marathon” and the “real world record” tells you one of two things about the future of the marathon, depending on your perspective.

Option one is that Kipchoge is just that good a marathoner that he has shown that it is truly possible.  “No limits”, as Kipchoge himself said.  If you lean this way, then you would expect a genuine sub-2 to happen soon without the carefully designed strategy we saw in Italy.  Kipchoge is indeed the world’s dominant marathon runner, having now won seven of his eight races, including the Olympics, London and Berlin, scaring the World Record once already.  If anyone was going to move the event forward, it was him.

Or, option two is the realization that some of Nike’s tactics were so effective that they were worth between two and three minutes to Kipchoge.  If you believe this, then the answer to the question “Where to next?” is in fact that we’ll return to 2:02:30 or thereabouts, just a shade under the current world record, and actually get two to three minutes slower for the foreseeable future.  It means that unless Nike (or someone else) designs another staged event like this one, that sub-2 hour barrier will remain safe for a while longer.

Telling these apart requires that we unpack a little bit about what we saw from Kipchoge and Nike on the weekend.


The drafting effect – worth 90 seconds? More?

My opinion is that the single most effective tactic used on Saturday was to have Kipchoge run most of the distance close behind a Tesla vehicle onto which had been mounted a large wind shield pretending to be a clock.

There was no reason for the size of this clock, or the close proximity of the car to the runners, other than to block air.  They even provided lasers to show runners exactly where to run for maximum benefit.  It was in fact the most egregious and obvious “cheat” compared to a normal marathon that Nike used in Monza.  I’d have no major issue with humans pacing and sheltering another runner, but this was, in my opinion, a step too far, and worth most of the improvement we saw.

The benefit of this protection from the wind cannot be understated, but it’s tricky to calculate with great precision.  It’s certainly not as large as you’d get on a bicycle riding in a large peloton, but it is not insignificant either.  It is estimated that a runner at 2-hour marathon pace spends around 15% of his energy overcoming air resistance, and so reductions in that would be expected to produce time savings by virtue of the freeing up more power to actually run.

Quite how much is the question.


The estimate of drafting benefit

Have a look at the calculations in the table to the right, which I’ve created based on some great stuff provided by Dan Bigham in a highly informative Twitter exchange the other night (follow Dan on Twitter, here)

What Dan did was to estimate the time saving, in seconds, for a runner going at elite marathon pace, with the assumption that 20% of the cost of overcoming air resistance can be eliminated.  I’ll try to break down the process for you:

You start with the time and speed.  For my calculation, I’ve taken a guy running a 2:02 marathon, which is a speed of 5.76 m/s.

Next, you need an assumption of what the power output is.  This doesn’t make a huge difference to the overall outcome because it’s all relative anyway, but Dan has used 5.5 W/kg and a mass of 55kg.  That gives you the runner’s power output in Watts (302.5 W).

Now, part of that power output is “running power”, and part of it is “aero power”.

The aero power component is calculated as: P = 0.5 x Air density x CdA x Velocity3.  For the conditions in Monza, Air density = 1.239 kg/m3, which I’ve used in the calculations.

Using a CdA of 0.4 (the same as an upright cyclist, so a reasonable assumption, I think) gives, in this case, an undrafted aero power output of 47.5 W. This is the work required to overcome air resistance in a non-drafted scenario.  Note that it is 15.7% of the total power output.  The remaining 84.3%, or 255 W, is the run power in the undrafted situation.

Next, we have to make an assumption about how effective drafting is at reducing that “aero power” component.  For this, Dan has assumed that 20% of the aero power can eliminated by drafting.  That means the 47.5 W comes down by 20%, to a new value, called the “Draft aero”, of 38.0 W.

That means that the draft run power is now 302.5 – 38.0 = 264.5 W, and increase of just under 10 W for running compared to the undrafted scenario.

Next, we work out a ratio of the two run powers, again using that relationship between power and the cube root of velocity, and that gives us a “Run co-efficient ratio” of 1.013.  This is the factor by which the velocity would increase in this 20% drafted scenario.

It means that the velocity of 5.76 m/s would now be increased by a factor of 1.0123, to a new velocity of 5.83 m/s, thanks to an increase in the run power component when drafting.  That in turn means a new marathon time of 7231 seconds, or 2:00:31, and a time-saving, in this scenario, of 89 seconds over the marathon.

Here’s the big assumption though – how much does drafting save relative to being completely undrafted?  In the table above, the assumption is 20%.  To give you a comparison, a cyclist in a peloton saves 40% to 60%, so this 20% assumption is quite conservative, but probably reasonable for a runner behind a large group of other runners.

So what I’ve done next is show you how the time saving would look if the drafting was even more effective.  The graph below shows the expected time saving for a range of drafting efficiencies.  Here you can see that if you assumed 40% saving (line D), then the time reduction in the marathon would be 2:53.


The “conservative” assumption and an estimated saving of 1:30

So which is likeliest?  I don’t know with certainty, so I won’t venture a specific answer, other than to say that in the normal marathon scenario, where you have a pack of elite runners accompanied by three or four pacemakers for 25km, I’ve seen estimates that if you extrapolate that draft benefit up for the entire distance (rather than 25 to30km), it’s worth around 1:00, or 10% to 15% draft efficiency (indicated by line A in the graph).  A2 indicates the likely time saving in reality, because pace-setters usually finish their work by 25km, sometimes 30km.


The pacing and drafting paradox

While on this issue, there is what I think is a really interesting draft paradox when you look at the pacing efforts in world record events.  I left it out when I first wrote this post, thinking it would be over-elaborate, but it’s led to some discussion and confusion (including by me, I confess), so I’m adding it in now.

Yesterday, on Twitter, Alex Hutchinson raised a study where the drag co-efficient had been shown to be reduced by 85% through “perfect drafting”.  This co-efficient is part of the calculation for the “aero power” explained above.

The implications of an 85% reduction in this value are bizarre – it leads to an 85% reduction in “aero power” because they’re proportional.  And I’ve not shown this level of draft efficiency in the graph above, because it’s almost twice what you get in cycling and thus I believe unreasonable for running, but you can calculate that the benefit would be 5:50 for a 2:02 marathon runner.

Even if you scale this back because drafting behind other runners is never perfect, let’s assume it is half of that supposed benefit – 40%, which is shown on the graph.  Now the benefit is around 3 minutes, or 1:30 per half-marathon.  What’s the paradox?  Well, consider that in the last six world records, they’ve run more or less even pace, so a first half of around 62 min, followed by a second half of 61:30 – 62:00.

But that second half is achieved, largely, with no drafting whatsoever (0% to 5% draft efficiency, at best). So you have a first half, with some drafting, in 62 min.  Second half, no drafting, in 61:30.

If you accept that draft efficiency is 40% (much less than the supposed 85%), then you are satisfied with the implication that the runner is running two halves in more or less the same time despite having an advantage between 1:30 and 2:30 on the first half.  No chance.  The same is true for track world records, by the way – drafted for the first half, undrafted second half, and they don’t slow down.  That should make you question the benefit of the drafting.

Therefore, I don’t think that the benefit of drafting behind other runners is as large as it has been made out to be.  Certainly, what Nike did on Saturday with the wedge would be more efficient than anything ever seen before, but I cannot see how it can provide anything remotely close to an 85% benefit.  The implication is ridiculous.


Nett benefit 2:00, compared to Berlin 1:15

Overall, I think a reasonable assumption would be to suggest that running behind six pacesetters and the Tesla for 35 to 40km would be worth between 20% and 35% draft efficiency, which is less than cycling, but significantly more than typical drafting in running events.  And this equates to an advantage of between 1:28 and 2:32 (shown by Area C in the graph, or the midpoint of this, 2:00).

Certain imperfect assumptions must be made, but in order to be conservative for the sake of a beneficial performance assumption for the runner, you’d say it was worth 2:00 (20% to 35% efficiency), and the projection would be that Kipchoge’s 2:00:25 was worth around a 2:02:25 had he not been able to rely on the car, shield and six runners for basically the entire race.

Of course, you now need to factor in what would normally be provided in a race like Berlin, and this is the point shown by line A2 in graph – the estimate is 45 seconds, and that’s being incredibly generous, as I’ve explained above.  If you add this back, then the performance improves from 2:02:25 to 2:01:40

The end result is that comparing Monza to say, Berlin, is probably worth 1:20 to 1:30, and would suggest that Kipchoge’s time is worth around 2:01:40.  The rest would be made up of all the other tactics.


The shoe – 4% or a lot less?

The next avenue that Nike exploited was the shoe.  This received much publicity in the build-up to the run, because they had controversially claimed that it gave athletes a 4% advantage.  They even incorporated the 4% into the name – the Vaporfly Elite 4%.

I still believe that the best policy for the IAAF to adopt on this is to ban the inclusion of any devices that may act as springs, and to regulate the midsole cushioning material, precisely because it’s impossible to quantify a performance advantage of this kind of tech (the same logic applied to Pistorius – you do it on principle, not performance).

Anyway, on this shoe, I think it’s safe to say that the it is not worth 4% – if it were, then we’ve have seen some eye-popping performances because that shoe has now been used by enough runners over the last twelve months that we’d know.  Remember that 4% for an elite male marathon runner is about five minutes.  It would be obvious, a bigger effect than doping, if this is what all runners were getting.

That’s not to say that the benefit is zero.  I suspect that there is a small benefit, though here, we have neither data nor theory to even allow an estimation.  A previously claimed shoe advantage by adidas was that their Boost midsole was worth 1% to efficiency, and around 1 minute in the marathon.  Here too, data on performance was lacking, but it probably wasn’t quite 1%.

Is the Nike Vaporfly 4% worth an additional 0.5% to 1%?  Possibly.  Runners believe so, Nike’s marketing team and scientists do (no surprise there).  There is also some extra fine-print that some runners may be getting a little more, while others get nothing – having responders and non-responders to technology is common.  I know that previous iterations of shoes incorporating springs have been highly variable – some people get worse, some get a lot better.

Theoretically, if a runner gets a 0.5% advantage, then it plays out in much the same way that the air resistance advantage would.  They can run at a given speed using less energy, or could use the same energy to run that little bit faster, this time perhaps 30 to 60 seconds in a marathon.

The short answer here is that we simply don’t know if Nike’s shoe is worth any time at all, but reports from athletes and researchers says it is, and so this may have helped Kipchoge slightly.


Hydration and fuel

Every runner knows that getting their in-race fuel strategy right in the marathon is key.  This is even more crucial for our Ultramarathons here in SA, and so few South Africans would doubt the benefits of optimizing this.

I’m less convinced that it is a major source of advantage to most of these elite runners, though, because they are mostly getting it right already.   There are not many instances of elite marathon runners, particularly when breaking world records, encountering any such limit to either hydration or fuel.  Given that most of the marathon world records have been set with faster second halves than first halves, and without any “wobbly parts” where they slow down dramatically, it would seem safe to suggest that running out of energy late in the race is rarely a problem.

So while I wouldn’t dismiss it out of hand, and it certainly is a key factor in runners who are not getting it right, I don’t see much benefit over and above a normal marathon being gained by fancy carbohydrate drinks and free access to them.  If this is worth 15 seconds in a marathon, I’d be surprised.


The course

The selection of the course is Monza was primarily based on minimizing the number of turns and its pancake flat gradient.  There’s no doubt there is a direct time saving when you can eliminate 90 degree turns – for you and I, running at our speed, corners and intersections don’t matter, but at 21km/h, you’re looking at perhaps a second (about 6m) every time it happens.

Monza was perfect in this regard – no corners, only sweeping bends, so that may have been worth around 20s in a marathon compared to most courses (London in particular suffers from many sharp corners).


Other tactics

Then there were a few other strategies.  We heard much about the training support provided, but I find it difficult to believe that anything changed in the athlete preparation as a result.  It was claimed that Lelisa Desisa had changed a good deal in his based on feedback, but then he had the worst day of the three runners, finishing in 2:14.  Performance is complex, so knowing the effect of a general or systemic intervention at the level of training is just about impossible.

So it remains an unknown, but I think there is an arrogance to sports science that assumes it can improve on generations of wisdom and knowledge that the east Africans have acquired about how to prepare for the marathon.  Certainly, I doubt whether Kipchoge got any benefit.

Finally, they had arm-warmers, and supposedly state of the art clothing, and even wore stickers to reduce drag on their legs.  If this was worth even two seconds in a marathon I’d be surprised.  They got the swoosh on display though.


Overall conclusion – slower next, in the absence of similar contrived events. Even then, maybe

I’d conclude that the drafting effect of a car plus six runners provided the greatest benefit, in the region of 2:00 if you make what is a safe assumption.  The course was likely worth up to 30 seconds, bringing a conservative estimate to 2:30 in total.  If you want to be more generous, this would increase to 3:00 because of hydration/fuel benefits.

We simply don’t know the effect of the shoes, but if that is worth even 30 seconds, then we have Kipchoge getting a total benefit of 2:30 – 3:00 (excluding hydration, which I think is really trivial).  More if you think the shoe is that effective, less if you are skeptical.

It means that his 2:00:25 is the equivalent of a 2:03:25, less whatever you think normal drafting is worth.  Perhaps 2:02:40, which you’ll note is a lot like the current World Record.  That is some food for thought – for all the excitement about making a huge leap forward and redefining what humans are capable of in the marathon, his performance may end up reflecting what humans have already shown they’re capable of for the marathon!

It means, in conclusion, that Kipchoge didn’t so much break through the barriers of human endurance as he partly bypassed them.  He did not get close to a sub-2 because he overcame the potential limits of energy cost of running that fast – rather, those potential limits were shifted slightly, moved aside compared to normal marathons, by the engineers and the Tesla, and then he took full advantage of that to get close to the summit.  It is, to finish the analogy, like getting so close to the summit with the benefit of extra oxygen.

A remarkable achievement, but not one that directly tells us anything about human potential in the ‘classic’ sense of what a marathon involves.

As for the original question – “what next?”- I think we get slower before we get faster.  I strongly suspect Kipchoge would, on a good day in Berlin, break the World Record, and could probably run 2:02:30ish.  That’s my pick for best case scenario in the foreseeable future, so I hope he runs it in good conditions in September.

That 2:02:30 is where we truly are in the marathon, without all the “aides” provided on Saturday.  If Kipchoge does that, then let’s talk about sub-2:02, which might happen within three or four more years.  But sub-2, legitimately, is still a way off.

If we see another marathon “trial/test” like Monza, then I think Sub-2 might be possible, given the combination of benefits at around 2:30. Whether Kipchoge can do it again, I don’t know.  It would have to be him, I think – there’s nobody else in the marathon who can, so it would be fun to see a similar event now, just to test that out.


P.S. A final word on doping, which absolutely has to be raised.  These athletes were tested after the race, and presumably on whatever normal schedule athletes in the elite testing pool would be exposed to in the build up.  However, for obvious reasons, there is very little confidence in that out-of-competition in Kenya, and we’ve seen, in the last year, that there is a problem in East Africa when it comes to doping.

Unless you believe that Sumgong and Jeptoo are the only high profile Kenyans who’ve doped, and that the testing has had a 100% success rate on them, there are without question other east Africans (and other athletes who train there) who are also doping.

I don’t see how anyone can have a conversation around the limits of human performance, with integrity, without at least acknowledging the reality that we cannot fully vouch for any athlete, especially those who are rarely tested out of competition.

In that vein, I really feel that Nike missed a trick by not prioritizing an increase in independent testing in the six months leading up to this trial.  Sure, it would not have been believed by many, but there were ways to go about it to at least provide some reassurance, over and above that provided by normal anti-doping (which is minimal).  Providing testing data at far more regular intervals, done by a trusted independent party, would’ve been a start.

In any event, the performance was spectacular, but it would be willful naivety to not at least raise this. We know there is an issue, we don’t have any specific reason to suspect Kipchoge or the other two, but pretending like this is not a possibility would be foolish.  I’ll just conclude by saying I really hope it’s a clean performance.

Do NOT follow this link or you will be banned from the site!
%d bloggers like this: