Marathon Splits

Most articles I write on Fetcheveryone come about because of something that went a bit wrong. Abingdon Marathon 2018 was my fourth marathon, and I would have been delighted to get round with an average pace of just inside 9:44/mile for a slender PB.

People often talk about a negative split in a race i.e. running the second half quicker than the first. This did not happen. I ran the first half at 9:34/mile, and ran-walked the second half at 10:49/mile. Or in percentage terms, I spent 47% of my time running the first half of the race, and 53% of my time running the second half. The real slow-down came in the last quarter though:

Arguably the cracks were showing in Q3, but maybe I'd created the problem earlier on by running the first half a bit too fast. And before we push on, it's worth noting that I was not just 'tired', but also feeling quite a bit of pain caused by a tight IT band. But this slow-down raises a few questions:

1. Is slowing down common amongst marathon runners?
2. Are faster runners better at avoiding a fall away in pace?
3. Is any failure point more pronounced in runners who start too quickly?
4. How many runners achieve the illustrious negative split?
5. How does the race outcome relate to long run pace?

We have lots of data on Fetcheveryone. As I write, 6561 Fetchies have recorded a marathon time in their Race Portfolio. But in order to make our data as relevant as possible, let's focus on the following subset of performances:

1. runners who have indicated that they were aiming for a PB
2. runners who have completed at least four previous marathons
3. runners who have completed at least four previous half marathons
4. runners with GPS data so we can extract accurate splits

This gives us 695 performances from runners who were giving it their all, with a reasonable amount of distance racing experience. Their times range from a 2:28 marathon down to a few six hour plus performances. Surely all these fine specimens would churn out negative splits! No. Only 23 (3.3%) of our runners managed to do this. The next graph shows the number of runners completing a negative split (to the right of the green bar), and a positive split (to the left). For example, 98 runners did the same as me, taking 47% of the time to run the first half of the race, and 53% of the time to run the second half.

Do we see better management of the two halves of the race from our faster runners? Yes. This graph shows our runners broken down by marathon time, and the median time spent negotiating the first half of the race. The faster runners are closest to achieving the negative split.

In my case, things didn't get tough until the last quarter of the race. And I'm not alone. A huge 622 (89.5%) of our sample runners ran slowest in the last quarter.

Let's consider the early pacing strategy. For each of our runners, I calculated their average long run training pace from their five longest pre-marathon runs. I wondered whether a big difference between long run pace and starting pace in the marathon would result in the biggest crash and burn.

This graph illustrates the number of runners and the pace they run the first quarter of their marathon. The bulk of runners set off between 5% and 20% quicker than their average long run pace.

Next I looked at whether choice of starting pace has an effect on the overall race outcome. For each runner, I calculated their pace in the first quarter (relative to their long run pace), and plotted this against outcome pace. The red line on the following graph shows each potential starting pace. The black dots show the median outcome pace for runners who started out at each pace. If negative splits were common, we'd expect to see at least some of those black dots below the red line.

Across the board, runners drop about 5 percentage points. For example, if you start out 15% quicker than your long run pace, you may end up with a marathon time that's 10% quicker than your long run pace. There are some hints here that if your starting pace is a good deal quicker than your average long run pace (say, more than 20% quicker), then you may be taking on a bit more risk.

Now let's get into some predictions. We've seen how the standard marathon predictor devised by Peter Riegel in the 1970's makes for some stiff predictions. But how many of our runners attempt to start at 'Riegel' pace, and how many are able to stick with it? I used the best half marathon times of our runners to make a Riegel marathon prediction for each of them.

212 (30%) of our sample set out at Riegel pace or quicker. Only 61 managed to stick with it - that's less than 9% of our runners. The majority of these also started at least 1% quicker than Riegel pace, and then managed to stay inside it. Only three of the 483 runners who started out slower than Riegel pace turned it around in the second half.

The following graph illustrates starting pace as a percentage of Riegel pace, and the outcome pace. 11 out of 20 runners (55%) who started out more than 5% quicker than Riegel pace held on to beat the prediction. 50 out of 191 runners (26%) who were less than 5% quicker than Riegel pace held on to beat the prediction. Only 3 out of 484 of runners (6%) who were slower than Riegel pace pulled it out of the bag to beat the prediction.

But we've moved on from Riegel here on Fetcheveryone. The Fetch Marathon Predictor offers a target that fits the data we've collected over the years. Therefore we'd expect to see a greater number of runners achieve it. By design, 77% of our runners get within 5% of our prediction (only 35% of runners get within 5% of Riegel).

The x-axis on the graph above shows the starting pace as a percentage of our predicted pace, and the y-axis shows the outcome race pace. The blue dots are runners who came in slower than the prediction. A considerable 160 of the 203 runners (79%) who completed the first quarter of the race 5% ahead of the prediction managed to hold on to beat it. 128 of the 322 runners (40%) who completed the first quarter less than 5% ahead of the prediction managed to hold on to beat it. And just like Riegel only 4 of the 170 runners (2.4%) who set out at a pace slower than the prediction managed to beat it.

Throughout this article, the data suggests that the vast majority of us should expect to slow down in the latter stages of a marathon - so IF you're desperate to achieve a time, it's worth contemplating a pacing strategy that takes this into consideration. But we're talking a percentage point or two here, and definitely not a crash and burn approach. The following table shows the paces you'd need to run a 49-51 positive split i.e. 49% of your time in the first half of the marathon, and 51% of your time in the second half. It's a delicate trade off between putting a few seconds in the bank here and there, and setting yourself up for a big L.

And just remember that the longer your race, the more opportunities there are for something to go wrong. A niggle can turn into an injury. A poor fuelling strategy can turn into an energy shortage. A negative thought can talk you out of believing. And there's stuff you can't control, like the weather.

None of our data takes into account any changes in terrain, hilly profiles, or those little moments of chance that can change your day. If there are big hills, off-road sections, or a midday weather forecast that you could fry an egg on, adjust your plan accordingly. Segment the race your way. Use our Pace Band Creator as a starting point, or make your own.

Ultimately, for all but the elite runners, time is not as important as having a good experience, and coming home safely. There's no shame in walking. There's no shame in stopping. Getting to the start line is way more than most people ever do. Don't be in too much of a hurry to leave it behind. Good luck!