Monday, 13 May 2013

The Negative Split Fallacy - Part 2 - The Explanation!


Yes, following on from lasts week's post where I used the data from this year's London Marathon to show that running a negative split is not an indication of a good performance, tonight I will hopefully provide an explanation into why the negative split doesn't work for non-elite runners.

In response to last week's post, and to previous pacing posts, I have received a number of comments.  Thanks for the comments, it is nice to know that people are reading these blogs, and the blog is achieving its aim, of getting people to question / evaluate their ideas on running, in this case on pacing.  Some of the comments I totally agree with, others have got me questioning why I believe my view that differs is more likely not to be wrong.

I guess the key issue, which many people have difficulty dealing with, is that if the negative split seems to work for the very very best in the World. i.e. the World recorder holders, then surely it should work for everyone.  Below are two comments received.
"As I pointed out in my recent post about the folly of positive split, elite marathoners achieve their best performances by running negative or very even splits. This evidence screams out the small negative or even spits are best, but you just dismiss that somehow Elite's and non Elite's are somehow different because they train differently, and that somehow "magically" positive split becomes best." Robert.
"No offence to hundreds of thousands of runners, but quite frankly I don't care that 95% of them run positive splits. The real question should be could they have run a faster time if they would have run more even splits? And again, who cares if it was even a PR - could that PR have been a better PR?  I think this topic is very hard scientifically to prove physiologically...but I do think a big stack of correlated data is useful. If you look at the data and nearly all world records at all distances come from even to slightly negative splits, that has to mean something. Random middle of the pack people running positive splits to me only solidifies the point." Brett.
How elite marathon runners manage to achieve a World record on a negative split, I don't know.  However, what I do know is that for non elite runners, attempting to achieve an even split or a negative split in a marathon is not the right approach.  I have provided the statistics to show this, e.g. 121 out of the first 122 finishers of the non-elite field at the 2013 London Marathon ran a positive split.  But now is the time to explain why for non-elite runners, the positive split is the best approach.

Firstly, the argument for the even split strategy as provided within a comment left:

"Mathematically and energetically, even splits are best. If you do a negative split, then you will have been losing time in the 1st half of the race. A positive split and you'll be losing time in the 2nd. Fact. Additionally, any variation in pace will generally result in poorer running economy. Of course, things are never as simple as that and tactics, course and environmental conditions will all have an affect on what is the ideal pace" Andy.

Yes, on the face of it, if we look solely at the energetics of running, keeping a constant physiological intensity does appear to be the correct approach.  The only problem is, that for non-elite marathon runners a constant running pace DOES NOT result in a constant physiological intensity.  This is where the confusion lies!  Now you will note that I again have used the term for non-elite runners.  This is because I have been unable to find published material within the scientific literature that has data on elite marathon runners, in terms of how much the physiological intensity increases throughout the duration of the race, whilst running at the same pace.  If anyone has got some literature on this aspect, please e-mail me the author names/title or provide a link within a comment.

So in paraphrasing/plagiarising Brett above:
"No offence to the elite runners, but quite frankly I don't care that the very very best of them run negative splits. The real question should be could I run a faster time if I would have not tried to run even splits? And again, who cares what the elite do if it was even a World Record - I am not an elite runner, I do not have the some time / resources / 'talent' etc,  what do I need to do regarding pacing so that my marathon time could have been a better PR?"

So what do I mean by a constant running pace DOES NOT result in a constant physiological intensity?  Lets look at some data.

First I will look at the following study: How Do Humans Control Physiological Strain during Strenuous Endurance Exercise? Jonathan Esteve-Lanao, Alejandro Lucia1, Jos J. deKoning, Carl Foster. This study monitored the running performance of 211 male middle and long distance runners with an average age of 32, and of various abilities. All trained for and entered competitions with the intent of achieving their best possible performances. Although not elite performers, all were serious competitors and some were successful in regional competitions.  The average time for the marathon was 3 hours 01 minutes.  The graph below displays how as the marathon race progresses, there is a continuous increase in heart rate, whilst at the same time there is a decline in running speed, starting at around the 20km mark, (ignoring the downhill section between 25-30km).

One explanation for increases in heart rate during endurance events is that it is a result of cardiovascular drift, and this is thought (not totally understood) to be due to a combination of dehydration, increased body temperature and increased sympathetic nervous system activity (e.g. adrenaline release).  Whether the percentage increase in heart rate attributable to cardiovascular drift truly represents the same percentage increase in physiological intensity is debatable.  Although the heart rate is increasing, the cardiac output (this being the amount of blood being pumped out of the heart every minute) may not be increasing, as the reason the heart rate has increased, is because the stroke volume (the amount of blood being ejected each heart beat) may have decreased.  I'm trying to stay away from the actual causes of fatigue tonight, so won't comment on the relationship between the heart rate and the rating of perceived exertion, and how this relates to fatigue, as I am trying to solely focus on the physiology.  But typically a rise is heart rate during a race means that the athlete is 'working harder'.

So if these 211 marathon runners wanted to maintain an even pace throughout the marathon, to avoid the decline in their running speed in the second half of the race, they would require an even more significant increase in heart rate during the second half.  Already the heart rate has increased from around 85% of HRmax until half way, up to around 90% of HRmax at the end of the race, so most likely an increase in HR of around 8 - 9 beats per minute.  To maintain an even pace for the second half of the race would require an even greater increase in heart rate, so as you can see, running an even pace DOES NOT result in a constant physiological intensity.

Just to get around the issue of whether the increase in heart rate does actually represent an increased physiological demand, if one wants to define physiological intensity as being represented by cardiac output rather than heart rate.  Then looking at how cardiac output increases during a race would be useful.  A recent study from 2012 did exactly this!  Using some new technology they managed to record cardiac output of 14 middle-aged runners running in a real marathon race, with an average finish time of 3 hours and 50 minutes (so a slower group of runners than the previous study).. The study was titled: Cardiac Output and Performance during a Marathon Race in Middle-Aged Recreational Runners. V´eronique L. Billat, H´el`ene Petot, Morgan Landrain, RenaudMeilland, Jean Pierre Koralsztein, and LaurenceMille-Hamard.

Again, as the marathon statistics from the London Marathon showed, with 96% of runners positive splitting a marathon race, there was a continuous decline in running pace as shown in the graph below.

The graph also shows how even though pace has significantly decreased, the cardiac output has stayed pretty constant, maybe a slight increase.  Again it is clearly obvious that if the 14 runners wanted to maintain a constant pace, i.e. run an even split, then there would need to be quite a significant rise in cardiac output.  The authors of the paper have illustrated this in the graph below, where they have expressed heart rate, stroke volume and cardiac output in relation to per metre run, so in essence in relation to running at a constant speed.
The continuous increase in all three physiological variables is apparent, therefore demonstrating that to run at an even pace DOES NOT represent an even physiological load, it represents an increased load.  The actual values per meter run reported were significantly greater for kilometres 36–40 in comparison to kilometres km 8–12, with increases of 21 ± 16% for Heart Rate Speed (Figure 4(a)), 25 ± 17% for Stroke Volume Speed (Figure 4(b)), and 26 ± 22% for Cardiac Output Speed (Figure 4(c)).  This data therefore shows how the one argument (apart from the World record argument) this being the so called optimal even constant pace strategy does not represent an optimal even constant  physiological load!

I appreciate that these 14 runners have an average finish time of 3 hours 50 minutes, so some of these runners were achieving slower finish times than what many runners aim to achieve, so some people may question the 'value' of this data.  In addition, the finish times had a standard deviation of 45 minutes, so a large range of finish times.   The fastest time was 2:57, and the slowest 5:10, hence why there are large standard deviation bars on the above Figure 4.  Taking into account this large variation in finish times, the researchers looked for any significant relationships between the physiological data and the marathon finish time.  The following graph displays one of the key relationships found.  This being that there was quite a strong relationship (r = −0.65, P < 0.01) between the upward drift in the Cardiac Output /speed ratio and finishing time.
The key is; what causes this upward drift in the Cardiac Output throughout the race?  If one is able to reduce this drift, which for these 14 runners was quite substantial, averaging an increase of 26 ± 22%, then there is likely to be an improved performance.  As illustrated in the above graph, the amount of drift does appear to influence marathon performance.  It is likely that as a runner gets physiologically 'fitter', one benefit is that there is a reduction in the level of upward drift in cardiac output.  And maybe the reason that the very best of the elite, i.e. the World record holders are able to achieve a World record on an even paced, or a slight negative split strategy, is that they are so physiologically fit that they do not have the same upward drift in Cardiac Output as the non-elite runner. Again, if anyone has go any HR / running pace data on elite athletes during a marathon please share it.

When comparing between the elite and the non-elite, one must also remember that the duration of the race for the elite (with a finish time of a little over 2 hours) is on average around only half the finish time of the average runner.  Therefore the average runner is running for double the amount of time!  As the upward drift in cardiac output is continuous over time, then the longer the time of the event, the greater the drift will be, and hence in terms of the optimal physiological energetics, the further away the runner must then be from trying to achieve an even race running pace strategy.

Now lets go back to the message I was trying to get across in my last post, based on the data from 25,000 marathon finishers (< 5 hours) at the 2013 London Marathon, although I only looked in detail at the the first 11,000 finishers (< 4 hours).  Still a pretty big sample size.  The table below lists the percentage average percentage slowdown for six finishing time bands.
So if we look at a 3:50 marathon finisher, the data from London Marathon 2013 indicates that on average they slowdown during the second half of the race by 9.9%.  The percentage increase in the cardiac output speed ratio in the above study was on average 26%.  They shouldn't necessarily be the same, i.e. the same percentage drift upwards of cardiac output being equal to the same percentage slowing down in running pace, however, one would expect the values to not be this different.  Why the difference between these two percentages then?  One is due to the way the percentages are calculated, where I calculated the percentage slowdown as the difference between the first 13 miles average pace, and the second 13 miles average pace.  Whereas the study calculated the difference between the fastest 4km split (8-12 km) and the slowest 4 km split (36-40km).

One must also look at the 14 subjects used in the study (middle-aged, male, recreational marathon runners all of whom were national-level short-distance triathletes or international level cyclists).  It is therefore likely that due to not being marathon runners, their legs may not have been 'conditioned' to the pounding they get whilst running a marathon, as opposed to cycling.  It is therefore likely that they may have experienced increased drift in cardiac output due to extensive damage and fatigue within the leg muscles.  I am only speculating here, but leg muscle fatigue has been shown to have a large influence on marathon running performance.  And with these 14 subjects being national level or international level, one would expect that aerobically they were very fit, so this excessive muscle fatigue could be a reason why the average finish time was only 3:50.

Anyway, regardless of the specifics of these 14 subjects, I hope that I have been able to illustrate above that in order to run at a constant even pace throughout a marathon, requires a progressive increase in physiological load, and therefore is contradictory to the even split physiological energetic argument often presented as to why the even split pacing strategy is optimal.

As I have just mentioned the issue of muscle fatigue, which has been shown is a key aspect with regards to marathon running performance, I'll expand of this aspect a wee it.  A recent 2013 study (Running Pace Decrease during a Marathon Is Positively Related to Blood Markers of Muscle Damage. Juan Del Coso et. al.) that investigated muscle damage in 40 amateur runners with an average finish time of 3 hours 12 minutes, found that their leg muscle power, tested immediately after the marathon, had decreased by 17%.  Interestingly the average percentage decrease in running pace was 16%.  So clearly with marathon running inflicting muscle damage, with the muscle becoming fatigue, there is obviously going to be an increase in physiological load as the marathon progresses, and hence the upward drift in cardiac output.

Again, a bit of speculation, but perhaps due to the extremely high levels of training that the elite marathon runners are able to perform, due to being full time athletes, then perhaps their legs are more 'conditioned' and therefore they do not 'suffer' the same amount of muscle fatigue, and so do not encounter the same upward drift in cardiac output.  Hence why it is not wise for the non-elite runner to try to 'mimic' what the elite runners achieve.  Hence why trying to run an even split / negative split for the non-elite runner is simply foolish!

In terms of how foolish.  What consequence will it have?  Well lets say for a 3 hour 50 minute runner. with a percentage slowdown of 9:9%, which is the average for a runner of this finishing time, then this runner on average would slow down 11 minutes and 23 seconds during the second half of the marathon.  So if they went out to go through half way in 1:55, they would miss their 3:50 target by on average a little over 11 minutes.  If they had just gone out that little bit faster, rather than running totally comfortable for the first half of the race, they would only need to have reached half way in 1:49:35, less than 5 and a half minutes quicker, and they would achieve their target finish time of 3:50.  By going out that little bit quicker, they are therefore able to slowdown their pace during the second half of the race, in order to keep the physiological intensity more constant , which is argued as the optimal strategy one should adopt.

Well, this post again has ended up being a bit longer than anticipated.  However, hopefully I have helped explain why the even spit pacing strategy is not wise for non-elite runners.  The fact that 96% of marathon runners positive split isn't because 96% of marathon runners are getting it wrong.  It is because the concept of the negative split / even paced strategy is WRONG!

I am suggesting that due to the on average 5.1 - 9.9% slowing down of pace during the second half of the race, for a sub 4 hour marathon time.  In order to counteract this, in reality this rather small amount of slowing down, the runner only requires a slight increase in running pace for the first half.  By acknowledging this percentage slowdown, which takes into account the physiological responses that actually occur during a marathon, rather than some non-evidence based ideal of an even paced strategy, then there will be an increase in marathon runners achieving their goals, which surely must be a good thing!

I will sign off tonight again with the quote from Tom from Marathon Talk:

“My money's still on the even / negative split but I'd be delighted to be proved wrong. My quote for the day... I'd rather know I was wrong than think I was right ;)" Tom Williams, 2011.
Hopefully after tonight's post, there is sufficient proof for Tom to stop encouraging the thousands of listeners to Marathon Talk to try to achieve an even split, or at worse a negative split.

All the best as you plan your pacing strategy for your next race,


PS I promise that this will be my last post on road marathon running.  I will get back to trail running blog posts, so apologies to those trail runners out there that aren't interested in road marthons, and the emphasis on mile split times and finishing times.

Yes, UltraStu will return to trail running, and just as a taster, I received an e-mail from the Cotswold Outdoor Store asking if I could promote their newly released trail running video titled "Why I Run: A short trail running film".  So to get back into trail running mode, here is the LINK to their video.  Take a look as I 'm sure it will remind you, after all of this road running talk (not that you probably need reminding), of the joy of running on the trails.


  1. Thanks for the additional discussion about HR drift/Cardiac output, all makes interesting reading and useful for debate of optimal pacing.

    Just like your analysis of percentages of negative splits in various races the data tells us some of the reasons why positive splits are likely to happen in races, and statistically just how often they happen. None of the papers or your analysis actually address the issue of what pacing strategy is optimal, it's only you that is applying this conclusion despite the nature of the data in front of you not supporting this or any conclusion on pacing strategy.

    To answer what is the best pacing strategy you need to do the different types of studies and statistically analysis. In terms of studies what you need to do is look at the effect on different splits on their effect on muscle damage/cardiac drift and other physiological markers. Without such studies we are left to speculate.

    With the statistical analysis you can't look at anonymous groups of runners in a single race. You have take groups of runners that you track over many races and study how their potential performances on the day may be correlated to their pacing strategy and the performance their achieve.

    One approach you could take is to look at splits for the PB's that runners achieve, looking at size and proportion of negative, even and positive splits. Then compare the probability of achieving a PB with any of these split types against the mass data like you have done. By looking at the relative probabilities you'll start to get hints at which split strategy is most effective.

    For instance in your analysis typically only around 5% of runners negative split, and if we found that only 1 in 20 people achieved a PB doing a negative split this would suggest that their is no particular advantage for negative or positive split. If less than 1 in 20 people achieved a PB doing a negative split then we could infer that perhaps a positive split would be more likely to result in a PB. If however more than 1 in 20 people achieved a PB doing negative split then we could infer that negative split runners are dis-proportionately more likely to achieve a PB than a positive split runner.

    If we look at the data for the elite marathon runners that I posted on my blog 2/3 used a negative split in their PB's. However, in an example like the London marathon none of male elits achieved a negative split and they all were off their PB's. For the elite's it certainly looks like a disproportionate number of PB's are done with negative splits and negative splits are key to achieving their best. The London marathon elite this year also suggest that they are far from immune of pacing errors and suffering the consequences - they are human too, if they go out too fast they suffer like the rest of us and in just the same ways.

    I hope this gives you an inkling of where you are going wrong in your analysis of the stats and where you could fix it.

  2. A second part of comment, needed to seperate as blogger doesn't allow for comments of silly lengths like mine :-)


    I'd also suggest that when thinking about the studies your discussed in this blog entry that you think more critically about what it might mean in terms of pacing. For instance these papers tells us that typically runners see cardiac drift, muscle damage and other physiological markers increasing through a marathon, even with pacing reducing. These papers don't tell us the reasons why, they just tell us what happens.

    The reasons are crucial for deciding upon pacing strategies. All the physical markers discussed in these are effected by running intensity. The faster you run the quicker you accumulate heat, dehydration, glycogen depletion and greater your rate of accumulating muscle damage. Just go run a half marathon a near PB pace vs marathon pace vs easy pace and you'll know that the physical stress you create isn't linearly related to pace, a small increase in pace has a disproportionate large increase in stress on the body.

    Our bodies have inertia in many of these physical aspects - it's takes a while for a body to warm up and begin to sweat, and with sweat we then later start to see problems with dehydration. Even when we slow down and stop generating so much heat it still takes a time for the body to home back to equilibrium - especially if compounded by dehydration.

    For muscle damage, once the damage is accumulated it stays with you for the rest of the race, it can't dissipate like heat.

    For glycogen depletion you can attempt to consume gels etc. to prop up blood sugar levels but you can never consume enough to replace everything that has already been consumed.

    I've only touched on a couple of these physiological factors and all of these cases going out faster will result in greater stress that has to be carried through the second half of the race and this is *exactly* what the studies show.

    What these studies show is the importance of avoiding accumulation of heat, dehydration, glycogen depletion, muscle damage. The best way I'd suggest to achieve this is an even or negative split. This applies to both Elite's and non elite's.

  3. Hi Stu,

    Read this and your previous blog with interest (well, you know I like looking at the statistics!). The figures are unarguable - nearly everyone slows down. It happens even more so in ultras, and the tendency for slower runners to slow down more than the elites is even more marked. Can't argue with that. However you sense there is a "but" coming, and this is it.

    I listened to Paula Radcliffe commenting on the difference between the men's and women's races during the London marathon coverage, her opinion was "you just feel so much better in the second half of the race when you have run an even pace, you don't feel that you are "hanging on"".

    Could this positive mindset you get have an influence on the outcome? I've run 15 or so marathons, some with positive and some with negative splits. I've always felt in better shape after the negatives than the positives. My best time was when I didn't look at the watch until halfway because it was a run I had decided not to take seriously, then realised I had more than enough left to speed things up. Similarly, possibly my best ultra performance (last year's Lakeland 50)came when I was able to run quickly in the last 10 miles when I would normally be very slow by that stage of the game, because I had decided it was a "training run" and gone very conservatively at first.

    You would say of course that on these occasions I would have gone even better had I started out faster; but the additional factor I am throwing into the mix is that feeling great over the second half of an endurance event compared with feeling stretched may have some effect on one's actual performance, for some people at least. (Or maybe not, that's the fun of an argument like this....)

    It would be interesting (but probably impractical) to see if there was any correlation between a runners pos/neg pacing strategy and how they felt at you need to finish absolutely knackered to have genuinely achieved your best on the day?

  4. October 2013 at 01:06

    couldn't agree more with Robert Osfield-all his comments, particularly about tracking a group of runners over different races with different pacing strategies. The fact that 95% of runners in a big city marathon positive split may just mean that 95% of runners have unrealistic expectations on how they will perform. In my experience, few of the runners I coach ever actually set out to positive split, they simply don't control their emotions at the start of a race, or are too inexperienced at the distance to really take on board my advice about being conservative a the start - they don't really believe that they will slow down as much as they do - I always find in subsequent races over similar distances their pacing tends to be much more even, taking into account conditions and courses. Also, the bulk of the field in most marathons and to a lesser extent half marathons, don't race over the distance that many times in their lives - they may be planning only ever to do one marathon and they set themselves a goal time before they even start into the bulk of their training. So, they may have a 3 hr 30 or 4hr goal, and they are going to either hit their mark or blow up trying. they don't want to be left wondering how they would have gone if they'd pushed it a bit more at the start. if you could measure these runners over time and using different pacing strategies, then you'd get some idea of the benefit (or otherwise) of even or neg splits vs positive splits

  5. +1 for the effectiveness of negative splits. I run all my races negative splits and I overtake so many people in the last mile or two it is incredible. All my PBs set with big negative splits. And I am nowhere near elite!! If 121 of the first 122 non-elite runners at the London Marathon ran positive splits then they simply started too fast and would have run faster with negative splits.

  6. "However, hopefully I have helped explain why the even spit pacing strategy is not wise for non-elite runners"

    The analysis on this is woeful and there are so many conclusions jumped to without any evidence or link:
    You argue in essence is the effort required to maintain a pace during a race increases over time which is obvious to most. That's definitely true, nobody would disagree.

    But from that you fail to even explore any relation of that to performance. There is no evidence here whatsoever to suggest that increasing the effort is detrimental to performance.

    "The fact that 96% of marathon runners positive split isn't because 96% of marathon runners are getting it wrong. It is because the concept of the negative split / even paced strategy is WRONG!"

    Again there's no evidence even supporting this in the entire argument true or not. You've just jumped at a conclusion. Just because 96% of people run a positive split does not mean at all that it's the correct way to run.