Weekly Mileage and Marathon Performance

A commonly held belief amongst runners is that increased weekly mileage will result in improved performance, especially at the marathon distance.  Several reasons exist for this belief, chief amongst them is the anecdotal observation that elite marathoners all seem to run relatively high weekly mileages.  In support of this belief, scientists have conducted numerous surveys of the training characteristics of marathon participants and detailed studies of the training characteristics of elite marathoners.  As a general rule, faster marathon runners tend to run more miles each week than do slower marathon runners.  For example a survey of 35 female marathon runners determined a fairly strong relationship (r = -0.74) between average weekly mileage and marathon performance (1).  Other surveys have found similar correlations.  These anecdotal observations and correlations are typically interpreted to mean that higher mileage is necessary for faster marathon performance.  While modern training theory does not suggest that mileage alone is responsible for running performance, it does place great emphasis on the importance of a relatively high volume of training.

While there may be a correlation between weekly training distance and marathon performance, it does not automatically follow that increasing weekly training mileage is the cause for the improved marathon performance.  As has been pointed out to me many times, correlation is not causation.  It is possible that weekly training volume exerts a strong influence on performance.  However, it is also equally possible that other factors are causing the improved running performance, but that weekly training mileage gets the credit.  Indeed, when scientists have conducted rigorous studies on the influence the various training variables have on performance, volume of training has not been shown to strongly influence performance.

So, just how influential is weekly training mileage on marathon performance?  Research by Dolgener and colleagues found no difference in the performance in novice marathoners who ran either 4 or 6 days per week (2).  Critics of this particular study point out that the total weekly training distance between the two groups was an average of just 9 miles and suggest that this difference is too small to reach a meaningful conclusion on the influence of weekly training mileage.  Therefore, lets take a look at two studies examining the influence of weekly training mileage and marathon performance.

Research

Grant et al

In 1982 a group of Scottish researchers wanted to examine the validity of the recommendation that marathon runners need to run relatively high weekly mileages in order to adequately prepare for the marathon (3).  To test the validity of this idea they decided to observe the training patterns of a large number of first time marathoners (but not necessarily novice runners).  The subjects of this study were 88 first time marathoners preparing for the 1982 Glasgow Marathon.  104 subjects began the study, but 16 dropped out due to illness, injury, etc.  The subjects were recruited from 225 runners who attended a training seminar for first-time marathoners at Glasgow University.  The subjects were self-selected in as much as they attended the seminar and decided to both run the marathon and keep precise training records of all the data, times, and distances requested at the start of the survey.  The average age of the subjects was 36.9 years, with a range of 18 to 70 years.

A number of variables were requested of each runner, including their marathon performance time and the average training mileage during the final 12 weeks of training.

The average weekly mileage of the subjects during the final 12 weeks of training was 37.2 miles.  However, the range of weekly mileage across the 88 subjects was quite extreme, ranging from 15 miles to 63 miles.  On the high end of the training spectrum, three runners averaged more than 60 miles per week for 12 weeks and eight runners averaged more than 50 miles per week.  On the other end of the spectrum, one runner averaged just 15 miles per week for 12 weeks and three runners averaged 20 miles per week.  So, while the average was 37 miles per week, the runners were not limited to a particular weekly mileage and the range of weekly mileage was quite large.

What were the results?  When correlated, there was a low -0.38 correlation between weekly training mileage and marathon performance.  What does this mean?  It means that there was a correlation between marathon performance and weekly training mileage.  As weekly mileage increased, marathon performance improved.  However, it was a very low correlation.  Weekly training mileage had a prediction coefficient of just 14%.  In other words, while weekly mileage had an influence on performance, it was a weak influence at best, predicting only 14% of marathon performance.  The researchers concluded that “…weekly mileage during training is a poor predictor of marathon performance.” and “There is no practical relationship between average weekly training mileage and race time, for novices at least.”

So we see that weekly mileage is a poor predictor of performance in novice marathoners and there is no real relationship between weekly mileage and marathon finishing times.  What about in experienced marathoners?

Franklin et al

Another group of researchers wanted “a) to determine the accuracy of predicted marathon finishing time among experienced and inexperienced marathoners, and b) to examine the relationship between training mileage per week and marathon performance”(4).  The researchers interviewed 158 contestants of the 1977 RWR Marathon.  Subjects were classified according to marathon experience as either first time marathoners, second time marathoners, or experienced marathoners (more than 2 marathons).  127 subjects completed the marathon, consisting of 63 first timers, 29 second timers, and 35 experienced marathoners.

The research showed that weekly mileage influenced marathon performance only slightly.  Correlation coefficients indicated that 16.7%, 28%, and 9.6% of marathon performance among first timers, second timers, and experienced marathoners respectively was explained by differences in weekly run mileage.  It was also discovered that for any given training mileage per week, experienced marathoners ran faster than did first time marathoners.  The results are tabulated in table 1.

Table 1: the percentage of marathon performance accounted for by weekly mileage

Runner Classification Influence of Weekly Mileage on Performance
First time marathoners 16.7%
Second time marathoners 28%
Experienced marathoners 9.6%

The researchers commented, “…the variation show(s)…that training expressed in miles/week (without control for specific frequency, intensity, or duration) offers little explanation for the differences in marathon performance amongst experienced or inexperienced runners.  Since distance is a function of intensity (speed), duration, and frequency, the present results highlight the inadequacy of training distance alone as a major determinant of performance.”

Discussion

Volume of training is one of the training variables and, as such, has an influence on performance.  Of that there is no debate.  The debate centers on just how much influence volume of training has on performance, with some placing great emphasis on it while others, myself included, place much less importance on it.  These particular studies reinforce the belief that volume of training does not exert nearly as much influence as conventional wisdom suggests.  With an average of just 17% of marathon performance being predicted by weekly training volume, other training variables obviously exert much greater levels of influence on performance than does volume of training.

In addition to the above findings, there is another item that is of particular interest from these two studies.  Marathon training programs in 1970s and 1980s placed their greatest emphasis on high weekly mileage, whereas today the emphasis is at least equally placed on the importance of a long run.  Grant et al noted that prominent researchers and running coaches of the day routinely recommended weekly training distances for the marathon of at least 40 miles per week and that runners were commonly cautioned that running less than about 45 miles per week would result in the runner “hitting the wall” during the final miles of the marathon.  Today, most training programs place at least equal if not more emphasis on the importance of the long run, with the recommendation that the runner increase the long run to about 20 miles three weeks or so prior to the marathon.

Assuming that the runners in these studies generally followed the conventional training advice of the day and focused mostly on total weekly training mileage, we could say then that the runners in these studies did not train specifically to run the marathon distance.  Instead, they trained to run a particular weekly mileage.  Removing the specificity of the long run from the training program, we are left with only the influence of weekly training volume on performance.  Therefore the results of this study are a direct reflection of the runners’ general training approach and are not influenced by the specificity of the long run.  If this is truly the case, then the marathon results of the subjects in these two studies were predominately influenced by volume of training and general fitness, providing strong evidence as to the true influence volume of training has on performance.

Recall the results of the marathon study by Dolgener.  In that study intensity and specificity were the same for both groups while weekly mileage was different for the two groups.  Marathon performance was equal for the two groups, with volume of training having no influence on performance.  As noted above, the difference in volume of training between the two groups was criticized as being too small to create a measurable result.  Considering the results of all three studies, the natural conclusion would be that volume of training does provide a small influence on performance and that relatively large increases in training volume are required before measurable differences in performance arise.

Summary

These two studies show that weekly training volume is a poor predictor of marathon performance and that volume of training exerts a relatively small influence on performance.  The results of these studies are in agreement with the marathon study by Dolgener, reinforcing the ideas that specificity and intensity are much more influential on performance than is volume of training and large increases in weekly training volume are required to affect small increases in performance.

Reference:

  1. Hagan R., Upton S., Duncan J., Gettman L., Marathon performance in relation to maximal aerobic power and training indices in female distance runners.  Br J Sports Med. 1987 Mar;21(1):3-7
  2. Dolgener, F., Kolkhorst, F., Whitsett, D.: Long Slow Distance Training in Novice Marathoners. Res Q Exer Sci. 1994; 65(4); 339-346
  3. Grant S., Sharp R., Aitchison T., First Time Marathoners and Distance Training, Br J  Sports Med.  1984 Dec; 18(4), 241-243
  4. Franklin B., Forgac T., Hellerstein H., Accuracy of predicted marathon time: Relationship of training mileage to performance, Res Q, 49(4), 450-459

Comments

Weekly Mileage and Marathon Performance — 26 Comments

  1. Pingback: Weekly Mileage and Marathon Performance | Training Science

  2. Pingback: Great Outdoor Blog » Blog Archive » Marathon Training: We’ve come a long way, baby!

  3. I completely disagree with your conclusion.

    If weekly training mileage explains around 15% of marathon performance, it doesn’t follow that other TRAINING facets explain the other 85%. For example, what if natural ability and/or body weight account for 80% of marathon performance? That would leave only 5% explained by other training aspects.

    If our natural ability explains a large portion of our finishing time, then weekly mileage could still be the largest manipulable factor in our training. There is a huge amount of research that needs to be done before weekly mileage is dismissed.

    • Research has shown that genetic factors exert the largest influence on performance – accounting for as much as 70% of an individual person’s performance. I reviewed that research in the 3 part series, How Much Can You Improve, posted on this web site. Natural ability, the term you used, is, by far, the most important factor in determining a person’s ultimate athletic success.

      The 30% or so of performance not determined by genetic factors is governed by other factors, the strongest of which is training.

      Within the portion of performance determined by training, volume of training is a significantly smaller factor than many suggest. That is the primary message of this and other research on this topic.

      • As i understand the studies posted in this article, training volume accounts for 15% of performance. If genetic factors account for 70%, then all other factors would account for only 15%. That would mean that training volume is by far the biggest controllable factor in running performance.

        Am I missing something? Perhaps the 15% for training volume is 15% of the non-genetic 30%, which would be 4.5% overall?

        • The studies didn’t account for differences in genetic talent. So, no, the 15% prediction power of training volume wouldn’t represent the 30% of performance accounted for by genetic factors. It’s not a simple matter of summing the genetic portion with the training and lifestyle portion of performance.

  4. One relevant performance-related variable that appears to be ignored is the age of the participants. It does not seem reasonable to me to lump every participant, across a wide range of ages, into the same group.

    For example, I highly doubt that it is fair to compare the relation between the volume of training with performance of a 20 year old against that of a 70 year old. What happens to these conclusions when you divide the participants into groups of 5-year age brackets (i.e., 20-24, 25-29, …)?

  5. Actually the studies cited here support the importance of volume in marathon training, not the opposie.

    The fact that results show that volume has a greater influence on less experienced marathoners than experienced marathoners and this actually means that volume is indeed an important factor in marathon training. Experienced marathoners leverage the physiology they have historically developed, whereas inexperienced marathoners rely much more on their direct marathon training for success. When you have less of a foundation built, that fact that volume alone can account for between roughly 17% and 28% of performance is quite significant.

  6. The correlation was significant. Don’t discount the effect simply because you had low factor loadings. I would have to agree with the above comments, that 17-28% seems like a large effect when considering training alone. I think you may need to adequately address mileage buildup in the Grant study. What about the correlation between VO2max and training mileage? Go back to the literature and look at important physiological predictors marathon time (VO2max, lactate threshold, running economy) and see how they relate in studies of training mileage. I do somewhat agree that mileage alone does not determine performance, but you simply haven’t provided enough of a comprehensive review of the literature. I could cherry pick 4 studies and write a paper on how we should only do HIT training because long runs damage the heart. However, the bulk of the literature paints a very different story.

  7. As with all these studies, unless all bodies in the test have the same diet, the same amount of sleep follow the same race strategy etc. etc. the tests have no validity and scientifically are worthless. They are then discussed as fact. Laughable really.

  8. While in some respects mileage in and of itself is not the supreme predictor of performance, it should be understood that some of the cell signalling pathways are time dependent and increasing mileage can increase the training stimulus. However, this is dependent upon the intensity being high enough. Ultimately, neither all slow long runs or all fast intervals will result in the best performance at all race distances. Some bleeding is required for vest results

  9. The conclusion of this article is just wrong. These are both poorly designed studies with lurking variables galore. As others have already pointed out genetics plays a very large factor in marathon times, so does age,weight, years of training, sex Male vs. Female etc. From my experience mileage is by far the biggest factor in our control. The longer the race the more important training volume becomes. For example a few years ago at the age of 33 I ran a 100 mile race while consistently running 30 miles a week. I placed 151 out of 218 finishers. The same year I ran two 5ks with equal to and double the amount of competitors and placed in the top 10 both races 17:31 and 17:34. Mileage matters. I’ve ran 3 marathons and they all highly correlated with the mileage I’d been running (weight varied 4lbs at most/age at race dates 26, 27, 32 so not much of a factor)
    marathon #1 3:48 = 16 miles per week training… Hit the wall at 18 miles
    marathon #2 3:21 = 30 miles per week…felt strong with plenty left at the finish
    marathon #3 4:12 = 0 miles per week…that was not a fun marathon. mileage matters a lot!

    • Hi, Justin.

      You claim that the studies were poorly designed with lurking variables galore. If you believe that to be the case, then I recommend that you apply the same standard to every training idea, belief, and anecdotal experience – i.e. reject any idea, belief, or training method not supported by well designed studies with no lurking variables. You claim that mileage is the biggest performance factor within our control, yet the data you provide to support your claim doesn’t even come close to meeting your standard for good data from well designed studies. Your only source of data is personal experience which is unreliable, untested, anecdotal data rife with lurking variables.

      Do you see the problem with the inconsistency of your argument? One could easily conclude that you hold a very high standard (well designed studies, no lurking variables) for data that doesn’t agree with your preconceived beliefs but will accept the most tenuous of observations, no matter how spurious they may be, as long as it fits your existing beliefs.

      If you really want to test your mileage theory, then you have to isolate one variable at a time and systematically test it. I suggest the following: Train for your next marathon by running 30 miles per week but do not run more than 5 miles during any run (i.e. no long runs). After completing that marathon and recovering sufficiently, train for another marathon by running 30 miles per week but include long runs (up to 22 miles in length). If you do this there will be one major difference between the 2 training programs – the length of the long run. While this would clearly be a case study of one, it would provide some valuable insight as to the relative importance of the long run on your personal marathon performance.

      Cheers,

      Rich

      • You are obviously more concerned with being clever than actually helping runners. Every person is different which makes a mockery of anything called Sports Science. In my opinion all studies supposedly based on any kind of so called science, when related to sport, should be prefaced by explaining that no two individuals are the same and that there is little therefore to be gained by taking it too seriously. If you do not think what I am saying is correct, read any coach online who will tell you that he has all the answers and you will find 5 others who say exactly the opposite.( eg. Fitz and Hanson brothers. Stress -recover stress again. Gradual building stress over time =residual fatigue.)

        • Andrew,

          The challenge with the argument “every person is different” is that it negates medical science. If every person is different then medical science couldn’t exist because the cure for any disease or ailment you can think of would be different for every person. For example, if we were all completely different then treating a headache with a single dose of aspirin would only work for one person. Every other person who ever got a headache would have to find their own medical treatment for that headache since a dose of aspirin would only work that one person. Everyone else would be so different that a dose of aspirin would not provide any benefit.

          The fact is we share more similarities than differences, which is why medicine exists as a science.

          The reason different coaches have different beliefs about optimal training is that, in the big scheme of things, we know so little about the bodies physiological adaptations to exercise. Exercise physiology is an infant science and there is much more we don’t know than we do know. In the absence of definitive research data, coaches are left to devise their own training beliefs and methods.

          • In the first part of your response you say how wonderful medical science is and in the second you say how little we know.
            Make your mind up. Personally I believe medicine is a calling and not a science.
            Presumably you are aware of the blunders governments have made in advising the things their populations should and should not eat on the advice of people who laughingly call themselves scientists?

          • Andrew,

            Medical science is not the same as Exercise Physiology. Medical Science is older, and much better researched, than exercise physiology.

            As for medicine being a calling and not a science, I ask – why not both? Can scientist be called to research medicine? Can Doctors rely on solid research in order to help them practice the art and science of medicine? When you get sick or need medical assistance do you go to a Witch Doctor or an actual Doctor. If your life hung in the balance, say following a terrible automobile accident, would you want the ambulance to take you to the hospital or to a faith healer?

            Finally, there has been an abundance of advice that has been handed out over the generations that has proven to be not be the best advice, or not based on actual research, or based on faulty research. But that doesn’t mean we should abandon science or research in an attempt to improve and advance our knowledge.

  10. It’s important to see the studies themselves and weed out slants in the presentation.

    You can view the first study Richard cited (Grant et al) at the researchgate site here, there’s a preview panel available:

    http://www.researchgate.net/publication/22401398_Accuracy_of_predicted_marathon_time_relationship_of_training_mileage_to_performance

    If the link above doesn’t work, google “Accuracy of predicted marathon time: relationship of training” and click on the researchgate result link.

    There’s a scattergram of mpw/marathon time for the identified 88 finishers on page 2. Take particular note of some of the relationships in the graph. For the runners at 50mpw and above (to ~63 mpw), the range was ~2:40 to ~3:40. For the 30mpw and below, the range was roughly 3:10 to 4:10 and a pair of runners at 5 hours.

    Also note the distinct vertical line of data points at 3:30. You can bet your boots there was a 3:30 pace bunny in this event. This feature skews some of the results.

  11. You might want to look at Tanda (2011) http://bit.ly/1ULoDei It clearly demonstrates that marathon performance is a function of both training distance and pace – both are important but not equally so. As to the correlation / causation debate, I am always surprised that there is any arguement. We know that fitness declines training reduces and rises with increased training. Fitness must be some function of the physiological stress applied. As a 2:57 marathon runner (49 year old male) I tested the formula and spent 8 weeks doing an average of 24km per day at 4:51 per km. I did no speed, hill or interval work – absolutely none. I did not even run as fast as marathon pace. Then at Frankfurt I did a 2:45:10 marathon – as predicted by the formula. We have tried a similar technique on about 10 other runners. They have all got very close to the times predicted by the formula. Racing speed results from the combination of training volume and training speed. But, it is incorrect to think that you have to run quickly to race quickly.

    • Christof,

      Thank you for the link to the Tanda study. Several things in that study caught my attention.

      First, the mean number of training days per week was 4, none of the runners in the group averaged 7 runs per week, and some ran less than 3 times per week (2.7 days). Tanda did not find a correlation between number of days per week training and marathon performance time. I can’t recall how many runners have claimed over the years that more days per week of running = better performance and that 3-4 days per week is insufficient for best performance. If I had to estimate, I would say that most serious marathoners I’ve talked to on running forums over the years have claimed that one must run 6-7 days per week to maximize performance. This study (along with others) does not support that belief. One can apparently run very well on just 4 days per week (remarkably similar to the results of the FIRST studies). Running more than 4 days per week is not correlated with better performance.

      Next, the highest correlation for marathon performance was the runner’s training pace. Those who ran the fastest in training (i.e. the fastest runners) also ran the fastest in the marathon. That should not be a surprise since marathon performance is highly correlated to performance at other, lesser distances. The guys who are fastest at 5K and 10K are also usually the fastest at longer distances such as the marathon. I suggest this is so because the same muscle fibers that enable one to run a fast 5K or 10K are the same muscle fibers that enable one to run a fast marathon, given adequate training.

      Third, and not surprising, adding in the mean training distance per week increased the prediction correlation. The average weekly mileage was about 41 miles, with a range of 25 miles to 69 miles. The average training distance was 9.6 miles per workout. All of these are within a range that I would expect to see, especially for faster runners like these (their range of marathon finishing times was 2:45 – 3:36) and not vastly different than what we have seen with the FIRST studies. One does not need to run a 100 miles per week to maximize marathon performance. In fact, if you set aside the 2 outliers, then there was essentially no difference in performance between running 37 miles per week (60km/week) and 56 miles per week (90 km/week). A 50% increase in weekly mileage (increasing from 60 km/ week to 90 km/week) caused no improvement in performance.

      Training and fitness can be graphed by an inverted U. No training = no fitness. As training volume and intensity increases, so too does fitness. But there is a point of diminishing returns where ever more training produces less and less fitness (and more and more injury). This particular study reinforces that idea. Unfortunately, it didn’t include runners running particular high weekly mileage so the study doesn’t definitively establish for this group where the peak of the upside U might occur. However, the performance of the group running 60 – 90 km/week does provide some evidence that about 40 miles per week may be close to optimal for runners with the talent to run a marathon in about 3 hours.

      I also found it interesting that the mean maximum distance run in a week was just 54 miles (range 37 – 82), which is about half the distance the “run 100 miles per week for best marathon performance” group advocates.

      • Dear Rich, I plugged my 2011-2013 data (distance run in 8 weeks and average pace) into the prediction equation and for the 9 marathons I did in that period (following conventional mixed marathon training) I got an r2 of 0.9 between the predicted time and actual times. I had marathon times between 3:30 and 3:07. So, the equation seems to work for me. I wondered if I might try a training technique of running perhaps an average of 173 km per week for 8 weeks as three almost equal distance runs per day – every day – at an average pace of 4:55 per km – probably in the 4:20-5:30 per km range. Of course with that high mileage I would do no fast running at all – perhaps nothing even as fast as predicted marathon pace and I could not afford any rest days. The Tanda equation predicts I should be able to run a 2:45 marathon at the end of the 8 week period. As a 49 year old runner who took up running in 2009 do you think this would work? I would love to get a 2:45 time. It would be the ultimate test of whether distance or pace is critical. Of course I wouldn’t want to do it if there is no chance of it working. What do you think? Worth a try?
        Greetings,
        Christof

        • Christof,

          I candidly don’t know if it would work for you. Some runners thrive on high mileage. Others, not so much. If your body can adapt and absorb that level of training,, then your performance will improve. If the training load is more than your body can handle, your performance will likely be sub-optimal. I do know that as weekly mileage increases so too does the incidence of injury. Still, the old saying is “no guts, no glory”. The only way you will know is to give it a try.

          • Dear Rich,
            You say weekly milage increases the incidence of injury. Very plausible of course. But would you (or anyone else) know a study which quantifies this a bit?
            Thanks!
            Maarten

          • Just found this. You really should make your mind up as you are saying exactly what you told me previously was silly in that we are all different. The most difficult thing in the world is to recognise the truth: even when it is staring you in the face.

Leave a Reply

Your email address will not be published. Required fields are marked *