Friday 29 January 2016

What makes you a faster runner - pace or distance?

Before I show what actually happens if one trains in the lands where 'dragons may lie' (long distance slow running), I thought it might be a good idea to consider what we know or think we know about those lands. Common phrases suggest that running high mileage at slow pace is not a useful strategy for a performance runner - Running slowly makes for a slow runner - Junk miles - Quality not quantity - Race pace training - Tempo running is a key session - He's a plodder - No pain, no gain!

However, we also know that elite runners engage in high mileage - or at least relatively high mileage - compared to most club runners. So, what does that training space look like when plotted on a graph?

In the figure below (just one - I am short of time this evening!) I have plotted three versions of training space (the 8 week average of training pace against daily distance run). All of the panels have the Tanda (2011) dataset plotted as filled circles together with a thick black line to the right of which there is no data - runners normally don't train with those types of averages. In panel A I have plotted shortened Tanda marathon prediction lines (i.e. without extrapolating the formula into areas where there is no data). This is a pure form of use of the Tanda prediction equation. We have no evidence that the formula correctly predicts data that extends beyond the dataset shown by the black circles. So, the space to the right of the thick black line is uncharted territory - and in keeping with tradition (we are in Cambridge!) we mark our map "Here be dragons". But, of course, most people believe that there is greener grass the other side of the line, just not quite green enough to be worth taking the time and effort to explore.

Figure 1. Training parameter space (8 week average pace versus distance) with three views of the unexplored area to the right of the thick black line. A, Tanda (2011) contour lines plotted approximately over the space for which data exists. Within this region the equation predicts marathon performance time reasonably accurately from average training data. B, The Tanda contour lines have been extrapolated beyond the dataset to longer distances in the general form that many people believe may represent the performance from training - running longer makes you faster, but not much faster. C, The Tanda (2011) equation extrapolated well beyond the original dataset - it is a mathematical prediction of a very different training load on marathon performance. The extrapolation contradicts notions of specificity and suggests that training load is a 'scalable' function.
In Panel B I have attempted to capture that in the form of modified Tanda contour lines. These coloured lines (in Panel B) are not a mathematical equation - although they could be - but represent the commonly held notion that running further does make you a faster runner, but not by much. So, the pink arrow in Panel B shows that a 3:00 marathon runner could train at 4:25 mins per km at about 11 km per day - or at 4:55 mins per km at 24 km per day. But, the extra mileage (twice as much) gets him/her no gain in performance just a potentially slower training pace. The obvious cost-benefit analysis - if one believes that the unexplored territory has performance benefits of that shape - is that adding on additional miles is of limited value: they are Junk miles. The underlying assumption is that pace is critical and that high mileage is just necessary for the fastest runners. However, the Tanda (2011) equation, when extrapolated into the dragon infested land suggests something rather different. In Panel C I have extended the Tanda prediction curves by simple extrapolation. The coloured lines suggest that a marathon runner of a given speed can train much slower if he/she runs further. Indeed, as the pink arrow indicates, a runner who covers a greater distance can afford to run slower and still show a performance improvement. Junk miles do still exist: if you run too slowly the all that happens is that you move along the same performance curve. I have calculated what pace would cause that to happen here.

The shape of these performance prediction lines suggest that the uncharted territory might actually be a rich source of potential improvement and personal bests - not dissimilar to Dick Whittington's streets paved with gold (helpfully protected by the 'Dragons'!).

But, to believe such an extrapolation one also needs to believe that the physiological systems which limit a marathon and are being trained are being done so by 'scalable' stresses rather than by highly specific types of training. The tempo run is no more important or effective than an easy longer run as far as producing marathon fitness - one just feels an awful lot harder than the other. This is somewhat heretical. Can it really be the case that a 5 km interval session produces no more benefit than a 20 km easy run? If so, it removes the focus from the leg muscles as the main system that is being trained since long slow running is a completely different stimulus to the muscles from fast running.

My feeling is that most people don't believe there are dragons to the right of the line - but, nor do they believe there is gold either - they see fatigue, injury, loss of form and speed: a pointless waste of time. Now, that is surely worth a test - time to train with the dragons!

Next: Junk miles


  1. Hi Christof
    Fascinating post - do you have any thoughts about the optimal distribution of training stresses across a week? Carl Foster proposed a simple method of quantifying training load (in 1998 I think), whereby session load was the product of session RPE (measured using the 0-10 scale) and session duration. These could be added to calculate weekly / monthly load etc. however there are a number of ways to generate the same weekly load, so Foster proposed the idea of training monotony (weekly load / SD of weekly load) and suggested that monotonous schedules characterised by little day to day variation in load resulted in less adaptations and increased risk of overtraining. A further variable could be calculated which is 'strain' (monotony x weekly load). It is thought that individual athletes have threshold levels of load / strain / monotony, above which they are at increased risk of lack of adaptation and overtraining. Anyway.... Based on these ideas , it would seem your proposals may work best by either keeping speed constant and varying daily distance, or else keeping distance constant and varying speed. Have you experimented in any way with different distributions of loading?

  2. You are absolutely right that different distributions of training loads within a 'cycle' could produce better results than the overall average suggests. We have noticed that runners who do 'race-efforts' within the training period tend to perform better than the prediction. However, we have not attempted to quantify this - it is quite difficult to do. Whilst high SD within training parameters (either speed or distance) is potentially beneficial, it also presents an increased risk of injury. This risk - and the actual injuries - often limits training resulting in poorer performances. So, whilst fast and very long runs are great for driving improvements, it only works for the few runners who can tolerate that load. I think this is where many runners get stuck. They attempt to get faster by pushing the intensity of individual sessions and train less as a result of the damage. One way around this is to vary the SD but keep the intensity equal. One can calculate how speed and distance might be traded with one another to do this. I don't think the Tanda formula is the best one to use for this since it describes cardiovascular fitness and not training damage. A simple product of distance x velocity^2 (ie kinetic energy) may be more useful. Kevin O'Holleran is experimenting with the idea of using training loads tuned to adaptation resonance to see if greater adaptations can be produced. I am still of the belief that avoiding injury whilst applying the maximum long term load is the critical feature......But, all of this is icing on a cake comprised mainly of volume and pace.

  3. This comment has been removed by the author.

  4. Hi Christoph, and thanks for some really fascinating posts. As someone who struggles with interval workouts due to injury, I'm particularly intrigued by the idea that you can run longer but slower and still hit the same target.
    Given your article and the Race predictor, I don't understand what's going on here. The implication is that it's possible to run at slower than your target marathon pace, and even for relatively few weekly miles, and still gain fitness. It seems too good to be true, especially when you consider that the difference in marathon time between 30- and 40-mile weeks at around 9-minute miles is just a few minutes.
    Assuming I haven't misunderstood it, what effect is at work here?

    1. You need to be a bit careful extrapolating the function outside of the dataset used to produce it. All of the runners were capable of doing a more-or-less flat paced marathon and could do a marathon between about 3:45 and 2:45. Outside of those boundaries your 'success' may vary. The formula simply quantifies training stress in terms of pace and distance. Speed and distance both drive improvements with a certain amount of cross-talk and you can get the same improvements as running faster just by running more. The effect that is at work is simply training - there is nothing new. By running you adapt the muscles, circulatory system, skin, energy turnover etc. By running further you drive similar adaptations to running faster. The cross-over may well be linked to something rather simple - one possibility is similar weekly amounts of exercise induced carbohydrate metabolism. Running short distances quick will burn through similar amounts of carbohydrates and running further at a slower pace. There are many other possibilities. But, it is not snake oil. Run more and you will generally perform better at the marathon. To run more you will almost certainly need to slow down. Of course there are also diminishing returns. Once you get over about 200km a week the benefits of more miles is pretty slim...

    2. Hi Christof, thanks for a very prompt reply (and apologies for mis-spelling your name first time around).
      Understood about the limits to the formula. It's nevertheless encouraging that an extra ~35% of moderate-paced running - even slightly *below* marathon pace - can result in the same sort of times as a 'traditional' marathon plan. I'll certainly be playing with this idea.