The sub-2 hour marathon attempt: The pacing strategy

So the weekend is nearly upon us, and the Nike attempt at breaking the 2 hour marathon will happen in Monza, Italy.

There’s obviously considerable interest in the attempt, both positive and negative.  I think for now, I’ve said what I think is negative about it, so let’s talk about the specifics of the race.  What is most interesting to me is that we’re going to witness a scientific experiment where a number of different interventions (shoes, pacing, hydration and fluid, route, scientific training) are thrown at three great marathon runners, and the hypothesis is that these interventions will add up to an improvement of 2.5% compared to what a human being has run before.

First, let me make a prediction, or my own hypothesis:

I don’t think all those tactics will be enough, unless they really do have some sorcery in the shoes.  I suspect that the combined benefit is at most 1.5%, mostly shoes and course, and so I suspect what will happen is the following:

They’ll hit the halfway mark more or less on schedule, in about 59:58.  Only two of the three will be there, Kipchoge and Tadese. Desisa will have fallen off at about 18km, I reckon.

By 25km, they’ll be slightly outside 2-hour pace, maybe 71:20 actual vs 71:04 target.  Then it keeps drifting slower, so they’re 45 seconds slow to 30km.  Then it comes down to what they’ve agreed.  If they’re incentivized to finish with big time bonuses, then I can see Kipchoge hanging on to run about 2:02.  Let’s call it 2:01:55.  Tadese around 2:05:xx.

If not, then I can see them shutting it right down and finishing in 2:06, or maybe even a DNF for all three.  The former would mean “jogging” the final 12.2km in about 40 minutes, but off that pace, might be normal.

So that’s what I’d predict, and it’ll be fun to watch.  What will be particularly interesting is how they go about pacing it and what it does to the athletes in the second half, which is what I’d like to talk about next!

What is the ideal way to pace the attempt?  Should they go out a little faster than 2-hour pace, and try to put some time in the bank for the second half?  Should they be conservative to begin with, and then look to finish very fast?  Or is it better to set that pace car at 2:50.6/km and hit every split bang on the desired pace?

It’s difficult to study pacing strategy in very long events. In short events, it’s a little easier because you can force athletes into different strategies, like starting 10% faster than some previous average, or 10% slower, and then seeing how they finish and what it does to the overall performance.  That’s ideally what you’d do to determine what is optimal, but it’s not so easy when the race or time-trial lasts 30 minutes or two hours!

We know from these forced pace studies that for short duration exercise (less than 3-4 minutes), the optimal strategy is a fast start, resulting in a positive split and a slow down in the second half.  As duration/distance increases, that changes, and an even split becomes optimal.

Let’s have a look at some data from previous world records suggesting this, starting with the 10,000m for men, because it’s obviously “immune” to route and profile variations.  This is the kilometer time, in seconds, of every kilometer from the 35 world records with timing:

What I’ve done is to divide the 35 world records into three groups, by era (n = 12, 11 and 12, respectively). That was somewhat arbitrary, but the principle it ends up showing is important.

You’ll see how the early generations (green and yellow) show a typical upside-down bowl pacing strategy – the start a little fast, then get slower in the middle, and then at the end they speed up significantly for the final kilometer.

The result of this, if you simplify somewhat, is that they run a slightly positive split (second half slower, but not by much at all):

1921 – 1953 = 14:52 and 14:56 (49.9% and 50.1%)

1954 – 1977 = 13:58 and 14:07 (49.8% and 50.2%)

Then the final era changes things, because that slowing down in the middle of the race is no longer present.  Instead, you have a fast start, fast middle, and then note that you still see an increase in pace in the final kilometer.

The end result is that that batch of records has a slight negative split (13:27 and 13:22, 50.2% and 49.8%).

Why am I telling you this?  Because it’s important to understand that this suggests that at the limits of human physiology, for long-distance events, running as close as possible to even pace is the best strategy for optimal performance.  If you could put time in the bank, then you’d see more world records set with faster starts and a slowing down at the end. You don’t.

In fact, only one world record in that last batch from 1978 to 2005 was set with the first 2km faster than the last 2km (that was by only 2.3s).  In 10 out of 11 WRs, the last 2km were the fastest of the race.

At the limits of human performance, even pace is the way to do it, with that small ‘endspurt’ in the final 10%.

Now look at the marathon.  Here are the half-splits from the last five world records, plus Bekele’s 2:03:03 from Berlin last year.  I’ve shown the times, and the proportion of total time for the first half vs second half.

Clearly, world records are being set with even paced strategies, at least when comparing halves.  Bekele’s 2:03:03 last year was run off the back of the fastest first half ever in a marathon – 61:11, and they paid for that somewhat in the second half to run the most “lopsided” positive split marathon of the six shown above.  That was despite a great race between Kipsang and Bekele, where Kipsang regularly tried to force the pace on the second half.

Of course, looking at halves only doesn’t really tell the story.  For instance, consider Patrick Makau’s WR of 2:03:38, compared to the WR at the time, by Gebrselassie:

So Makau, in green, might have run reasonably even splits of 61:45 and 61:53, but the manner of his race was far from even, with huge variations in pace from one 5km segment to the next – he dropped a 14:20 5km split from 25km to 30km, for which he paid later, running 14:59 from 35km to 40km.

Gebrselassie, in contrast, was never more than about 15s per 5km off the pace required for his WR – far steadier, and with the faster finish typical of long distance events.

Point is, you can achieve even splits in different ways, so we should really be talking about segment by segment pace.  Part of the appeal of this route, 17 laps of a 2.4km circuit, is the precision in pacing it allows, so we can do this with more resolution.  Here’s how I’d do it:

The table below is how I’d attempt to structure an attempt to run 1:59:56.  The overall pace is 2:50.6/km, giving 10km targets of 28:26 (quite frightening to think – only a few times in the history of the marathon have people even run 14:13 for a 5km split!)

What I’d do is go out slightly slower for 10km, then pick it up in order to be just outside at halfway, on course at 30km, and then I’d allow it to drift slower between 30km and 40km, before a fast finish to get in under 2 hours).

Not that I think this would work, mind you.  I reckon what is likely, as I said up top, is that they’ll hit halfway in around 60 minutes, and then I suspect it will start to drift slower.

Slowly at first, maybe by 2 seconds per kilometer, then faster, so that by 30 km, they’ll be around 45s slower than required.  Then it’s decision-time, as I said, and I suspect Kipchoge might be able to hang on for a 2:02, but more likely will get much slower and finish outside the current world record.

We actually have a really good illustration of what happens when a marathon runner goes out very aggressively, because that’s more or less what Mary Keitany did two weeks ago in London.  She went to 10km in 31:17, which is just a little slower than her half marathon PB pace from earlier this year.  Admittedly, the first 10km in London are slightly downhill, but that helps the illustration, because what you’ve got is Keitany, running slightly downhill at a pace that is slightly slower than her half marathon PB, for the first 10km of the race.

This is basically exactly what Nike will have the three do on the weekend, except they won’t have a downhill profile to assist them – they’ll have other “aids” like pacing and shoes.

But look at what happened to her race after her blistering start:

She got slower and slower, eventually running a pretty substantial positive split.

And she provides the illustration that I think will play out for Kipchoge on Saturday.  He’ll go to 10km in 28:26, and maybe to halfway in 60:00, which is slightly slower than his half marathon best (59:17 in Monza just seven weeks ago).

Then comes the slow down, and the question is whether he hangs on to run under the World Record, or whether he falters so badly that he misses even that?

Remember how pacing strategy is achieved – the runner is constantly assessing their physiological state against some known endpoint and their own expectation of the effort.  Motivation obviously plays a huge role, along with the physiological information coming from various systems in the body – how hot am I, how much oxygen do I have, what is my metabolic state etc.

At some point in the first 20km, that feedback will overwhelm the system, because they know that they have to sustain a half marathon pace for twice that distance.  They may be able to push through, to hang in there for a while, but I reckon physiology will win the day by around 30km, and the pace will have to drop significantly.

If Kipchoge can get to 30km in around 1:25:30, and then hang on to run the last 12km at about 3 min/km, then he’s going to run 2:02, which is my prediction for the perfect day.

I do feel there’s just as much chance that he doesn’t finish it at all, but it all depends how aggressively they go with that pace car, and how Nike accommodates the situation if they start to drop off.  Will the car keep going at 21km/h?  Will they slow it down to 20.5km/h as early as 15km?

Time will tell, literally.  It should be interesting, and I’ll post more thoughts here in the coming days.

Ross

The Science of Sport - Scientific comment and analysis of sports and sporting performance