In this post, we will discuss the methods to analyse resisted sprint mechanical outputs, and detail a simple, direct method to determine the entire load-velocity relationship and the training parameters that have a kinetic backing applicable to the acceleration phase. As a typical example of illustration/application, we will discuss the estimation of the load (sled mass in this case) that allows athletes to produce their maximal mechanical power output (Pmax) over a few seconds in sprinting, i.e. with a 50% decrease in the theoretical maximal velocity “V0”.
I warmly thank Matt R Cross, Pierre Samozino, Pedro Jimenez-Reyes and Ryu Nagahara for their contribution and help with the content of this Article.
Basics and concepts
In the approach we present here, the horizontal component of the ground reaction force (that corresponds to the force produced by the athlete backward, in the horizontal direction) is computed according to a macroscopic model. Because this method has been validated during overground sprint running, using the most accurate method available (force plate system), we can be confident that when applied correctly the data from this method are representative of what truly occurs during sprinting (acknowledging a small magnitude of error, and its ‘macroscopic’ nature). In this macroscopic approach, that informs on the sprint output “big picture”, “mechanical power” is considered as the product over time between this force output and the displacement velocity of the athlete’s center of mass in the anterograde-posterior direction. To read more on how force, velocity and power output are computed in the sprint context, you can read this narrative review of literature.
During a “free”, unresisted sprint acceleration, the velocity, horizontal force and mechanical power output follow these typical changes over time (example here from Usain Bolt’s world record race):
These figures clearly show that Pmax is only reached around 1 second and the athlete spends a very short period around these ‘maximized’ conditions. Thus, adding resistance to a normal sprint is a way to maintain the athlete in conditions around Pmax, which itself is only possible by producing a maximum effort against a resistance that will allow the athlete to reach Pmax at the velocity plateau at the end of the acceleration phase, and so at a maximum velocity (Vmax) that is equivalent to the velocity at Pmax. This is what we call the optimal velocity (Vopt), and two experimental studies by our group showed that this maximum velocity is the half of the theoretical maximum velocity “V0”. Please note that this general concept theoretically applies to all parts of the force-velocity spectrum. For example, the kinetic conditions specific to high force (e.g. the first step or two of the sprint) are only maintained for an instant in a “classical” unresisted sprint, except for the maximal velocity plateau during which velocity is approximately constant. We actually validated this theory in our recent paper, which showed that the kinetic conditions experienced at maximum ‘resisted’ velocity (i.e. at maximum velocity per the loading protocol being towed) are similar to that experienced during the corresponding part of the acceleration phase.
The method that we will detail in this article is a way to determine that individual Vopt, and the load associating this Vopt to the velocity plateau in a simple way, without computing friction, which might be complex in practice.
What the F…..riction !?
When adding load onto a sled and pulling it during a sprint acceleration, in addition to an increased mass of the system being accelerated (i.e. body-mass plus that of the sled and the resistive force generated in the direction of running is equal to the weight of the sled+load (mass times gravity) mediated by the friction coefficient between the sled and ground surface (and other potential factors, such as speed). This coefficient might be accurately determined using several devices (see our paper on this topic here), although all have their limitations. However, due to reasons of complexity, and expense associated with assessing these parameters accurately, practitioners may wish to forgo these computations altogether. In any case, considering the effects of friction during resisted sprinting (particularly in sled sprinting) this is key, since the same given load (e.g. 20% of the athlete’s body mass (BM)) on different sleds (new, old, rusty) and different surfaces (dry track, wet turf) might show very different friction coefficient. In fact, during the initial tests of the friction coefficient of the sled that we used in our ‘optimal’ sled loading study were troublingly un-reliable – until we realised that our readings were changing substantially (i.e. halving the effective resistance experienced by the athlete) due to the paint wearing off the bottom of the new sled. And so, we argue that, except in strictly similar friction conditions, expressing loads in %BM is inaccurate when directly inferring the resistive force that will be induced. For example, in our study we report ~80%BM is optimal at approximately 0.4 value for the friction coefficient, but this could feasibly be much greater on other surfaces.
For research, unless directly measuring the ground reaction forces of the athlete there is no choice if one wishes to compute the horizontal force production of the athlete. However, for practitioners we think that a “velocity-based” approach to resisted sprinting, in the case of impossible friction estimations, is the preferable approach. Indeed, finding the load that will induce a Vmax decrease of a targeted velocity (e.g. 20 or 50%) may be performed on any type of surface conditions, and provide accurate assessment regardless of whether the actual friction is known. The sentence “you should work at the load that will be associated with you reaching 80% of your maximal speed” is an overall more accurate approach than “you should tow XX% of your body mass”. The former instruction does not depend on the friction conditions (or to be more accurate takes them into account) whereas the second one does.
The simple method and associated spreadsheets
This method is based on 2 steps, for which we have designed 2 spreadsheets. It is “velocity-based” since the objective will be to determine Vopt first, and then determine the load that will lead to the individual reaching a Vmax equal to their Vopt. The main advantage of this simplified approach is that you won’t need to measure the sled-ground friction coefficient.
STEP 1- From an unresisted 30-m sprint, and using the split times (timing gates or MySprint app), you can determine the sprint mechanical profile, including the maximal theoretical velocity V0 (velocity axis intercept of the force-velocity relationship). This can be done entirely with the app, but our current practice is to use the app for identifying the 5, 10, 15, 20, 25 and 30-m splits as in the following video (using this free spreadsheet):
Note that a recently published study showed that this analysis can be performed from only 10, 20 and 30-m times, provided the start of the timing system is concomitant with the very first propulsive actions of the athlete. Any delay between these two moments will lead to substantial errors.
STEP 2 – Our theoretical and experimental studies have clearly shown that the velocity at which maximal power was produced was equal to the half of V0 (see above). In addition, Vopt as estimated using V0/2 during a single unresisted sprint was shown to be close to that determined with a multiple trial protocol with various loads (see theory and experimental studies).
Note that our results technically apply to the entire FV spectrum. The aim of this step will be to determine the individual relationship between sled load and the Vmax reached at the end of the acceleration. Our observations consistently show that this sled load-Vmax relationship is linear, so once it is determined (from 4 sprints as explained below), the load that is associated to any velocity (e.g. vopt, as V0/2) can be easily calculated.
Perform 4 maximal ~30-m sprints (or more/less depending on the distance your athlete needs to reach top speed)with enough rest in between to fully recover (5 mins). For the first sprint, we always perform an unloaded sprint (i.e. with zero loading). For the remaining sprints, the idea is to select a relatively even span of loading that will allow you to see >50% decrement in your unloaded maximum velocity in the last load. You can monitor this decrease live, with each sprint, and make calculations on the fly. Since the relationship is linear, if the athlete performs a good trial the spacing of loading and magnitude is not super important. From our work, typically 25%, 50% and then 75% of your body mass (in total, sled mass included) has been sufficient, but you will need to check this specific to your surface (some athletes and surfaces have needed >100% BM to go past a 50% decrement in V0). You may shorten the sprint distance in the heaviest conditions if you obviously decelerate past a given distance. Film the sprints with MySprint and report the 5-m split times in the spreadsheet. Note that if you have 6 pairs of timing gates you might enter the gates times instead. The 100% body mass sprint is not necessary since this 5th sprint data will align with the 4 other sprints and the linear equation will be similar, so you might skip this one.
Then, use this free spreadsheet and follow the instructions in this video:
Our observations show that when sprinting with the optimal load obtained with this procedure (for instance after a couple of minutes rest after the 4 first sprints), individuals reach a Vmax that is really close to their “Vopt”, i.e. their V0 divided by 2. We would recommend performing this quick test as it is a very good way to double check that the load you have determined here is the load that will induce this Vopt. Our research and practice with athletes from various sports showed that, this load was highly variable among individuals (from less than 50 to more than 120% body mass on a dry synthetic surface) and it is of course only valuable for a given athlete per each sled-surface condition. Any significant change in training surface or sled must lead to re-assess this optimal load.
No load versus light versus high load: what mechanical outputs?
This section is based on a paper from our group that is currently in review (request for pre-print here).
We explained the concerns we have with these studies in a commentary paper and a Letter to the Editor that are currently under review. The primary problems we see with the current narrative around the computation of forces and practical applicability of approaches are the misunderstanding the methods currently proposed (for example, the ‘velocity-based approach’ we describe, and validate in our work), and the publication of results based on inaccurate (and at times somewhat non-transparent) methodological processes. In fact a systematic review of the literature published falls prey to both of these problems, and appears to base a substantial portion of their conclusions and practical recommendations on that in the same issue of the International Journal of Sport Physiology of Performance than our paper detailing the current approach, that seem to closely match equivalent force plate analyses (study in preparation).
The aim of the commentary in-review is to provide a clear narrative of the measurement and computation associated with assessing horizontal force in resisted sprinting, since this information helps us gauge the application and usefulness of loading parameters. In the manuscript we use the methods of two recent papers published (our own, and the study of Monte et al.) as examples of the difference in findings with methodological errors and different ‘models’ of interpretation. To summarise: computational rules need to be followed since the accuracy of results hold obvious importance, and the model of looking for ‘instantaneous’ increases in a targeted variable (e.g. power during the acceleration phase) may not be advantageous because of the minimal time spent in this condition during the acceleration phase.
One key point is that the load that is pulled during resisted sprints influences the Vmax reached, and thus the mechanical power output produced during the Vmax plateau. To show this we willuse the example of an athlete that we tested (running velocity measured with a radar and data smoothed with an exponential function). In the figure below, you can see in black the power output as a function of time computed using our approach versus the approach of Monte et al. for a 20% body mass and a 80% body mass load. These computations are based on the equation below, that follows basic principles of physics, and have been shown reliable in an experimental study comparing these computations to reference force plate data.
For clarity reasons, in this example we used a constant coefficient of friction of 0.4, which is an average value for the conditions tested (see this article for detailed discussion), and we did not correct the horizontal component of the pulling force exerted on the sled for the angle of the tether. But these two simplifications apply to all conditions tested and do not change the story.
The computations proposed by Monte et al. give results that do not seem realistic: almost no force/power production at top speed during resisted sled sprinting. This computation error (we think) then leads to substantial errors throughout the acceleration phase, and certainly we suspect this may have effected the results. Once the plateau of velocity is reached, Pmax should remain ‘fixed’ in absolute value – since this represents the peak ability of the individual that (as observed in cycling studies).
The second and very interesting point in terms of training stimulus, is the total time spent at high power output (red zone in the figures below indicates >90% of Pmax). During a free sprint, Pmax is reached within the very first seconds of the sprint but this does not last long, because of the rapid decrease in horizontal force output (due to the increase in velocity). The main advantage of the optimal load is that it will force the athlete to spend much more time in the “optimal velocity” window, and thus work around Pmax for longer during a sprint. In the example below, the athlete spends less than 1 second at >90% of his Pmax in the free sprint, compared to more than 4 seconds when performing a sprint against his more optimal load. The lighter condition (20% body mass does not even allow him to reach more than 70 % of his Pmax.
To help your understanding, please try this simple exercise: pull a sled with 25% of your body mass at constant 10km/h and then do exactly the same thing with 75% body mass. In which case did you feel you produced more horizontal force?
When I speak to coaches about heavy resisted sprint training, I like to call it “the dish you avoid but have never taste-tested”: when I ask “honestly, how many of you have ever done a session with several sprints pulling sleds with more than 50-75% of your body mass”? The answer is never more than a few of them. While there appears to be a consensus that heavy sleds should be avoided (probably due to tradition, but also confusing research interpretations) the science is certainly not clear on this point: only one pilot study exists on the effects of training with very heavy sleds. As detailed in a previous post, the acute and training effects of this type of stimulus should not be mistaken. As such, coaches should be encouraged to be experimental and try and see for themselves.
The “negatives” of this approach are that of course you need to be fully familiar with the specific task of sprinting while pulling such heavy loads (this requires a progressive approach, but it is safe…most athletes squat/deadlift much heavier than they sled pull!). This method relies on the athlete being able to “extend” the maximal velocity phase in such conditions and theoretically ignores the substantial work required to reach this “Vmax plateau”, which might relate to fatigue factors playing a role in the adaptations observed.
Well, back to sprinting. With all that being said, it still seems reasonable to think that training in that “optimal velocity – maximal power” zone will lead to a more specific stimulus of the maximal power output both in terms of intensity (work close to Pmax) and volume (high time spent around Pmax). But what exactly do we know?
- Friction is complex, and requires costly and difficult methodological approaches to generate a relatively accurate picture of external resistive forces (paper here).
- We can accurately determine the kinetic output during resisted sprinting, and when considered at maximum velocity we know that the optimal load for maximum power corresponds to half of V0 – on our surface, this loading parameters was much higher than that proposed as optimal (loose terminology) for physical adaptations (paper here).
- The spectrum of force and velocity output from resisted sprinting can be applied to that during an unloaded sprint – allowing us to target particular phases of the sprint using loading external loading parameters (paper here).
OK, now what do studies say on training at this theoretical “optimal load”?
Training with the individual “optimal load”: does it work?
Haha…evidence! A challenge and big constraint we have is that people are eager for training evidence, but to have this we must have coaches and athletes who are willing to take a chance and step outside of the norm. From our point of view, coaches cannot simultaneously complain about researchers publishing data on non-elite athletes or doing no research in this area, and almost systematically refuse to have “their” elite athletes being tested (except some trailblazers we are super grateful to work with). The most frustrating part of this paradox is that our research does not require invasive, inconvenient or abnormal tests…they just need to sprint, which is supposed to be their favorite/everyday task.
So far, only two pilot studies investigated the training effects (i.e. not the acute effects) of resisted sprint accelerations with heavy sleds. The first pilot studyshowed greater improvements in maximal horizontal force and Pmax in amateur football players who trained with 80% body mass loads, compared to a control group who trained without load. However, this protocol used the same load for all players (80% of their respective body mass) and this might not have corresponded to their individual optimal load.
In a second pilot study, we determined the individual optimal load for the football and rugby players who participated in the study, using a protocol very similar to the one described above. Then, one group trained with that optimal load, and another group (randomized assignment) trained with a light load, i.e. the load associated with a 10% decrease in maximal velocity. The results of this study were contrary to our hypotheses since we observed that between pre- and post-testing (one week before and one week after the training program), both groups overall similarly improved their Pmax (trivial between-group differences)… no superior effect of training at the optimal load. As discussed in the paper, we have two main hypotheses for this unexpected outcome (following the analyses previously detailed here, we still theorise that training at the optimal load is a superior stimulus for Pmax and represents a balanced approach):
First, the training content was progressive in the optimal load group and they fully worked at their optimal load only in the last weeks of the program. We think that any further study on the topic should have the athletes working at their optimal load throughout the program duration.
Second, and this is our main hypothesis, we tested the athletes for unresisted sprint mechanics (including Pmax) at weeks number 1 and 12 of the program. Due to the important overload generated by training at the optimal (i.e. heavy to very heavy) load, and according to our observations on Pmax kinetics in rugby (see here), it is possible that the athletes in the optimal load group needed more time to adapt and benefit from the training effects, compared to the light load group. In other words, by testing both groups one week after the last training session, we might have missed the time of “peak power output” in the optimal load group, that might have occurred one or two weeks later, due to longer training adaptation kinetics. This idea came to our mind following discussions with the coaches who reported that some players in the optimal group were in really good shape a few weeks after the study, but not immediately after. This point is key since most studies on sprinting (and strength training overall) only consider a “pre-post” measurement approach, while heavy resistance sprint training might be associated with longer adaptation kinetics and justify an individualized “pre-peak” approach, i.e. repeated measurements over the 3-4 weeks following the intervention to observe the potentially longer and individual “time at peak power”.
Interestingly, this is what we wanted to test in a recent (unpublished) study performed by Pedro Jimenez-Reyes: 22 athletes trained for several weeks at their individual optimal load as determined in this article, and sprint measurements were performed one week before training (pre) and 1, 2, 3 and 4 weeks post-training. Their training content was normalized and controlled during this “post-intervention” period and did not include sled work. This allowed us to know (i) the time at peak power output for the average of the group (ii) the difference for the group between classical “post” measurements (1 week post) and the week at “peak” power output and (iii) the variability in the individual post-training adaptations: how many athletes had their peak power output at 1, 2, 3 and 4 weeks post-training? Very informative, and necessary study.
The main results are here:
Basically, the PRE-POST “classical” change in maximal power was 5.4% on average, with a large inter-individual variability (SD = 5.9%), which gives an overall small effect size. Most individuals (17/22) reached their maximal power output 2, 3 or 4 weeks after the training intervention, and the PRE-PEAK change (calculated from each individual peak value, regardless of when it occurred) was much clearer, and much larger: 10±5%. In other words, the “1-week post” measurement window is not appropriate in the context of high resistance sprint training (only 4 athletes out of 22 reached their highest Pmax at that measurement session). This is a very interesting and important track for further research on high resistance sled sprinting, we think researchers should not only investigate the PRE-POST changes, but also consider the clearly individual and delayed adaptations to heavy sled sprint training, and the PRE-PEAK changes. From an individual performance perspective, that’s what matters! Coaches seek to improve individual performance, not only group average, so research should focus on individual performances, not group average adaptations.
In conclusion, we think, as too often said but not done, that “more experimental research is needed”, not more meta-analyses. One last argument supporting the fact that we need to dig deeper, and better into the “optimal loading” possibility, is that several coaches reported anecdotal evidence, often with rigorous field approaches, and positive results. See for example the very detailed report by Cameron Josse, or the Altis report by Jason Hettler (Part I, Part IIand Part III). Let’s keep experimenting…
Cross MR, Brughelli M, Samozino P, Morin J-B. Methods of Power-Force-Velocity Profiling During Sprint Running: A Narrative Review. Sports Med 2017. 47(7):1255–69.
Cross MR, Tinwala F, Lenetsky S, Samozino P, Brughelli M, Morin J-B. Determining friction and effective loading for sled sprinting. J Sports Sci. 2017;35(22):2198–203.
Cross MR, Brughelli M, Samozino P, Brown SR, Morin J-B. Optimal Loading for Maximizing Power During Sled-Resisted Sprinting. Int J Sports Physiol Perform. 2017;12(8):1069–77.
Cross MR, Lahti J, Brown SR, Chedati M, Jimenez-Reyes P, Samozino P, et al. Training at maximal power in resisted sprinting: Optimal load determination methodology and pilot results in team sport athletes. PLoS One. 2018;13(4).
Cross MR, Samozino P, Brown SR, Morin J-B. A comparison between the force–velocity relationships of unloaded and sled-resisted sprinting: single vs. multiple trial methods. Eur J Appl Physiol. 2018;118(3).
Morin J-B, Petrakos G, Jiménez-Reyes P, Brown SR, Samozino P, Cross MR. Very-heavy sled training for improving horizontal-force output in soccer players. Int J Sports Physiol Perform. 2017;12(6).
Petrakos G, Morin J-B, Egan B. Resisted Sled Sprint Training to Improve Sprint Performance: A Systematic Review. Sport Med. 2016;46(3):381–400.
Rabita G, Dorel S, Slawinski J, Sàez-de-Villarreal E, Couturier A, Samozino P, et al. Sprint mechanics in world-class athletes: A new insight into the limits of human locomotion. Scand J Med Sci Sport. 2015;25(5)
Romero-Franco N, Jiménez-Reyes P, Castaño-Zambudio A, Capelo-Ramírez F, Rodríguez-Juan JJ, González-Hernández J, et al. Sprint performance and mechanical outputs computed with an iPhone app: Comparison with existing reference methods. Eur J Sport Sci. 2017;17(4):386–92.
Samozino P, Rabita G, Dorel S, Slawinski J, Peyrot N, Saez de Villarreal E, et al. A simple method for measuring power, force, velocity properties, and mechanical effectiveness in sprint running. Scand J Med Sci Sports. 2016;26(6):648–58.