Recovery Time: A New Variable

Figure 1

I recently came across a website I found interesting: Training Science. Normally, this is not the kind of resource to which I am attracted, it is neither a professional scientific study nor is it by an author I know and trust. The reason this site caught my eye is because the views it expresses tended to align with conclusions I had come to on my own. Two such conclusions are that genetics dominates everything else (including training) in determining athletic performance and that for most people, less is more in terms of training. In fact, agreement with my prior prejudices should have been a reason to avoid this resource rather than seek it out. The reason to avoid this site is confirmation bias, the tendency that we all have to seek out data that agrees with our prior beliefs and to avoid data that contradicts those beliefs when an effective search for the truth would have us do the opposite. Thus, I may be giving this site more weight than I should because it agrees with my prior beliefs. Despite that, I will proceed, but will do so with caution.

Training Science advocates for a number of controversial opinions. For example, it questions the importance of VO₂max and of lactate threshold for predicting real world athletic performance. It does provide legitimate, scientific references for many of its assertions, making it no worse than the newspaper reports that have been inspirations for a number of my blog posts, and as in those cases, I have followed up by reading the references and evaluating them critically. In this post, I am only going to discuss one of their iconoclastic opinions; that many people ought to be exercising significantly less often than what is commonly recommended, and the reference they cite to support that assertion.

This is the assertion from Training Science:

"[W]hen the subjects were training just 3 days per week it took an average of .9 of a day for the subjects to recover.  ...  recovery time increased to 3.6 days when they increased training to 5 days per week.  ... It would seem that even though the subjects were recovered enough to repeat a previous performance, they were not completely recovered ... [and] if you choose to workout prior to full recovery, it will take you even longer to recover.  If you persist in training prior to recovery, your level of fatigue will grow and along with the increasing level of fatigue will come a slower rate of recovery and a decrease in your rate of improvement (and I would add an increase in the risk of injury)."

The reference they cite in support of this assertion is "Effects of training frequency on the dynamics of performance response to a single training bout" by Thierry Busso, Henri Benoit, Régis Bonnefoy, Léonard Feasson, and Jean-Renélacour. J Appl Physiol 92: 572–580, 2002. Figure 1, shown at the top of this post, comes from that reference, and illustrates the important numbers Training Science would like to use from this study. Specifically, t_n = the time it takes to recover sufficiently from a ride so that you have the same performance as you had before the ride, and t_g = the time it takes after a ride to maximally develop the increase in performance that the workout provides. This figure reminded me of so many diagrams I had seen in so many training articles; this is accepted dogma in the exercise community as to what the training response looks like. Virtually all such articles then go on to recommend that, after a workout, one should wait for a time t_g between one workout and the next, so that each workout can "build" on the increase in performance from the previous workout. At first, I thought the scientific paper referenced by Training Science was a simple quantitation of this conventional training advice. When I looked closer, things got less clear, and by the time I finished evaluating this scientific reference, I felt like the authors at Training Science didn't really understand this very complicated paper, and that the paper said much less of practical value than Training Science thought it did.

Critical Analysis of Busso et al.

This study involved six subjects, a very small number. They were men in their 30s who were not participating in an exercise program prior to the study. In the study, they first exercised for 8 weeks, Monday, Wednesday, and Friday, took a week off, and then exercised Monday, Tuesday, Wednesday, Thursday, and Friday for four weeks. Exercise involved riding a stationary bicycle. During the study, the maximum power the subjects could generate was measured on Monday, Wednesday, and Friday. As one would expect, that increased with exercise. This is the raw data from the study:

Figure 2

This figure could not be simpler. A-F are the six subjects, and each maximum power measurement is plotted as a function of time after the start of training. Maximum power (which I refer to as performance) is the highest average power the subject could maintain during a five minute test. The training program is diagrammed on panels E and F; LFT is the 8 weeks of training 3 days/week, HFT is the four weeks of training 5 days a week, with one rest week in between. Just looking at this raw data, there are two obvious stories one could tell. The story Training Science would have you believe is that the rate of increase of performance went down when training was increased from 3 to 5 days a week. This story is perhaps best supported by subject D, for whom a plateau in improvement seems to begin at about the same time as HFT. However, there is a second story, offered as an alternative by the authors of the study, who note that in any training program, progress plateaus, there is only so good an athlete can get no matter how long they train. This story says that there is not much difference between training 3 and 5 days a week. Rather, there is just the natural plateau in improvement that occurs after several weeks of training. This alternative story is perhaps best supported by subject F, for whom the plateau in performance begins well before the initiation of HFT. The authors confess that a limitation of their study is that the High Frequency Training occurred immediately after the Low Frequency Training, and so the first impacted the second, allowing for these two alternative stories.

The above two stories are both based on the raw training data and do not provide nor use t_n and t_g, the recovery times shown on Figure 1. Where do t_n and t_g come from? they come from a mathematical model that the authors of this study and others developed over the years. To model figures like Figure 1, they use this equation:

Figure 3

p_n is the maximum power a subject was able to generate on day n which is determined by the amount of exercise done for each day i, starting on the first day of the study, up until the day before day n. p* is the the maximum power that subject generated before training, w is a function defined in the paper, and the remaining values are the parameters whose values are varied to make the model fit the raw data for each subject, shown in the Figure 2, by adjusting them using a best fit algorithm. Once they had the value of these parameters, they used them to calculate the recovery times t_n and t_g  as follows:

Figure 4

I confess I have not "checked their calculations" and thus am unable to independently evaluate their model. That is the bad news. The good news is that I do not believe it is necessary to do so in order to evaluate the relevance of this paper to development of a practical training plan, an evaluation I will describe below. That said, here are the results of those calculations:

Figure 5

The X axis on these graphs is the day after start of training, and the Y axis is the calculated recovery times. The authors of this paper suggest that t_n is the time it takes after a single workout for performance to return to the value it had before that workout. Training Science uses the increase in t_n during weeks 10 through 13 as evidence that training 5 days a week is worse than training 3 days a week.  Are they correct? What should we make of this data? How should we decide?

Let's start with a reality check. Note that Figure 1 does not describe the experiment done in this paper, the results of that experiment are shown in Figure 2. Rather, Figure 1 is a "cartoon", a visual description of what the authors suggest their results might mean. There are two important points to understand about Figure 1. First, that it is an assumption of the authors that the t_n they calculate corresponds to t_n shown on Figure 1 and that the t_g they calculate corresponds to the t_g shown on Figure 1. Second, even if the authors model is completely correct,  Figure 1 is not drawn to scale. As drawn, Power is shown to increase by about 20% after a single workout. Assuming their model is correct, the actual amount of increase after one workout is closer to 1%. Assume that t_n and t_g are what the authors think they are. Note that t_n is about one day, an entirely reasonable number. Looking at Figure 1, it would appear that t_g would occur about 2.5 days after a workout. However, t_g calculated by the authors is much longer, about 10 days, an entirely unreasonable number. Does it seem likely that performance will continue to increase for 10 days after a workout? If we assume the consensus of the training community is correct, that the time between workouts should be t_g, it certainly does not seem reasonable to assume that one should exercise only once every 10 days!

The authors provide a third reality check. If it is "bad" that t_n increases from about 1 day to about 3.5 days when training is increased from three days per week to five, then we should find that serious athletes have training schedules that match their measured value of t_n. In contrast to this expectation, the authors reference a previous study in which a value of 23 days was measured for t_n for one such athlete. Based on that, the authors suggest that, rather than suggesting a value of frequency of training, t_n might relate more to the time that a successful athlete should rest (taper) between a training regimen and an important competition. This either means that the conventional wisdom of the exercise community, that the interval between training sessions should never be less than t_n and ideally should be t_g, is wrong, or else the t_n and t_g calculated in this paper do not correspond to what is shown in Figure 1.

The Bottom Line

I do not believe that the reference cited by Training Sites supports their argument that increasing training from 3 days a week to 5 is harmful. I think the reference they cite is of significant long term theoretical interest, but at this stage of its development offers little of practical value to an athlete. Do I disagree that 3 days a week is better than 5? As it happens, I neither agree nor disagree, I think "it depends", but that is just my opinion, I know of no scientific study supporting or refuting that assertion. I think the guidance offered by the exercise community, which includes Training Science (once you understand that Science has little to do with their common sense recommendations) can provide a good starting place and valuable sanity check for designing a training program. At present, I see no evidence that the scientific community has much to offer in the way of practical training advice, though I believe progress is being made and that could well change in the future. Finally, I believe we are all different, and any training plan has to be tuned to our specific set of interests and capabilities, and in the end, I am left with the only guidance I have found useful for such tuning, listening to my body.