Lose less to win more: A comprehensive examination of the velocity loss literature
The velocity loss literature provides valuable insight on how far from failure the majority of training should occur, but it isn’t as straightforward as one might think. Join me as I address how to integrate the velocity loss data into the proximity to failure data in order to provide the optimal RPE ranges depending on the percentage of 1RM on the bar and the number of repetitions that are performed within the set.
- Velocity loss is related to neuromuscular fatigue; thus, training at lower velocity losses (~0 – 25%) may be optimal for strength adaptations
- The velocity loss data should be integrated into the proximity to failure data in order to provide generalized optimal RPE ranges
- The optimal RPE ranges are specific to the percentage of 1RM on the bar and the number of repetitions that are performed within the set
Non-failure training has been supported to have a slight edge over failure training for strength adaptations (1). However, a frequently asked question is, “How far from failure should I train?” Thankfully, I’ve been able to dissect the velocity literature in order to reveal the exact answer to you, but it isn’t as simple as proximity to failure a is superior to proximity to failure b. Rather, the optimal proximity to failure (or optimal RPE ranges) is contingent upon the percentage of one-repetition maximum (1RM) on the bar and the number of reps that are performed within the set.
In this article, first I’ll provide a brief summary of the research on velocity loss (with a brief intermission on some important statistics that are sometimes interpreted incorrectly). Second, I’ll explain what the velocity loss literature is suggesting, how I’ve determined the optimal proximity to failure, and why we must integrate the velocity loss data into the proximity to failure data. Finally, I’ll finish up with some practical programming strategies specific for strength athletes.
A Brief Summary of Research
As of today, there are currently 7 longitudinal studies investigating different velocity loss thresholds on strength; however, there’s been 5 published this year alone, so I’m sure that we can expect many more in the near future (2 – 8). In all 7 of these studies, only one study has shown a significant difference between groups for 1RM strength (8). Specifically, over the course of 8-weeks 25% velocity loss (VL25) resulted in significantly greater 1RM strength adaptations compared to VL50 in the body mass prone-grip pullup (8). The main findings of each of the 7 studies is illustrated in table 1.
Table 1. Velocity Loss Long-Term Training Studies
In all 6 remaining studies, all groups significantly increased 1RM strength; however, most importantly, there were no significant differences between groups. Despite this finding, this data still provides some useful information in my opinion, which I’ll address after we take a short intermission so that I can provide a very brief lesson on statistics to understand how we can interpret these findings.
A Brief Intermission on Statistics
First, let’s begin with percentage change. In the case of these studies, we are specifically looking at 1RM percentage change from pre- to post-test. Percentage change is almost always reported, and something that is sometimes incorrectly referred to in order to “support” claims. However, percentage change is probably the least important statistic to report and if one looks at the percentage change alone without first looking at the p-value of a group by time interaction and the between group effect sizes (between group effect sizes are very important) one can make some wild and false conclusions. To provide an example, perhaps group 1 increased 1RM by 15% and group 2 only increased 1RM by 5%. On paper, this looks to “favor” group 1. However, if the groups are not counterbalanced for 1RM strength at baseline, these results don’t provide us with much useful information, if any at all. For example, if group 1 had a 1RM squat of 100 kg and group 2 had a 1RM squat of 115 kg at baseline, group 1 was less trained and therefore had the potential to increase 1RM strength to a greater magnitude than group 2 in the short term.
For this reason, we must look at the between group effect sizes in order to quantify the magnitude of difference between groups and determine if the change is meaningful or not. It is for this reason that you’ll always see effect sizes reported in meta-analyses (typically a Bayesian forest plot of between group effect sizes), and almost never see percentage changes reported. Based on Cohen’s D, effect sizes can be classified as trivial (<0.20), small (0.20 – 0.49), moderate (0.50 – 0.79), and large (≥ 0.80). There are other methods of classifying effect sizes, but I’ll keep it simple for this paper.
Between group effect sizes are particularly useful in exercise science resistance training studies that have very small sample sizes, in which sometimes meaningful differences (based on effect size) can be present despite no significant difference between groups (based on group by time interaction). On the other hand, in certain fields in which sample sizes are astronomical, sometimes there will be a significant difference between groups (based on group by time interaction); however, the difference is non-meaningful (based on effect size). Lastly, group by time interactions are also important to indicate whether there was a significant difference between groups or not, and also displays the probability that the difference is due to random error. Furthermore, Bayesian inference is used in order to provide the likelihood of the hypothesis being true or not, but this is typically never used in exercise science studies. In short, if the p-value is less than 0.01, you can be confident that this was in fact a true finding. However, if the p-value is close to but still less than 0.05, the results may not be a true finding. For all of the aforementioned reasons, this is why it is particularly useful to look at between group effect sizes.
A Brief Summary of Research
Now that we’re done with our brief intermission on statistics, let’s turn our attention back to the velocity loss literature. In the 6 remaining studies, the between group effect sizes for 1RM strength were all below 0.20 (trivial) except for in one study: the study conducted by Pareja-Blanco et al. in 2017, which is arguably the most well recognized of all the velocity loss studies (3). This one was a goodie. Not only did they look at 1RM strength changes, they also looked at hypertrophy adaptations via muscle biopsy (which is perhaps the best method to look at hypertrophy, yet it is pretty uncommon for the most part in resistance training studies). In this study, the between group effect size was 0.34 in favor of VL20 and the 1RM % changes were 18.0 and 13.4 for VL20 and VL40; respectively. Furthermore, the percentage of myosin heavy chain IIX was preserved in VL20; however, it was reduced in VL40 (group x time interaction: p = 0.04. However, again, with this high of a p-value, this result may be a “false positive”). Finally, there was no significant difference in hypertrophy between groups.
However, what’s interesting to note is that in the Pareja-Blanco et al. (2020) study, VL20 resulted in a significant increase in hypertrophy relative to baseline, whereas VL10 did not, even though there was no significant difference in the total number of repetitions performed between VL10 (143.6 ± 40.2) and VL20 (168.5 ± 47.4; 6). Therefore, this suggests that perhaps there may be a minimal VL threshold of ~20% required in order to induce significant increases in hypertrophy? Conclusively, there isn’t really that much else notable to report in all of these studies when solely looking at strength and hypertrophy adaptations. However, a final important point is that volume (via total repetitions) is un-equated in all of these studies; therefore, it is difficult to discern whether the results are due to the actual velocity loss itself or due to the differences in volume between groups. Of course, it’s probably a combination of both; however, this is still a limitation as it would be ideal to have volume controlled (and everything else controlled) in order to solely investigate velocity loss.
Overall, based on the totality of the velocity loss literature, I would very very cautiously state that there may be a small benefit towards training somewhere in the 0 – 25 VL range for strength adaptations, simply because training above ~25 VL doesn’t seem to provide any added benefit, but I think that we need more research in this area (particularly with volume equated studies) before arriving at concrete conclusions. In short, 0 – 10 VL is definitely not worse for strength adaptations compared to 20 VL, and I’d argue that 1RM in 0 – 10 VL would most likely have a meaningful between group effect size (probably just a small between group effect size) if volume was matched compared to the higher VL thresholds. Furthermore, I don’t see any added benefit towards training at VL thresholds above 25, since there is no significant difference in hypertrophy between VL20 and VL40. Furthermore, training at VL30 and above may induce unfavorable neuromuscular adaptations and result in failure training, which we know is sub-optimal for strength (1, 3). Finally, I like to provide +5% above the VL threshold of 20, simply because if you perform a first repetition that is slightly faster than normal, this can make the VL higher despite the set still being at the same proximity to failure.
What is this Data Suggesting?
First, I want to state that these studies are comparing specific VL thresholds to one another. Most of you are probably like, “Yeah, of course they are; I already know that”. However, the average proximity to failure can be inferred based on this data. For example, the repetitions that were performed at 80% of 1RM by the VL10 group from Pareja-Blanco et al. (2020) was 2.5 ± 0.9; however, in a low-rep group and a high-rep group it has been reported that 4.8 ± 0.6 and 7.1 ± 1.3 repetitions can be performed, respectively, in the smith machine squat (9). Therefore, for example, if some individuals performed 3 reps in the VL10 group but can only perform 4 reps at 80% of 1RM, they would have been at a 9 RPE. Conversely, if some individuals performed 2 reps in the VL10 group but can perform 8 reps at 80% of 1RM, they would have been at a 4 RPE. In short, the within-group variation in proximity to failure likely varied drastically.
Of course, on average, the proximity to failure were most likely higher in the VL40 group compared to VL20, VL10, and VL0. But, instead of there being an optimal proximity to failure that is universal to all percentages of 1RM, perhaps the optimal proximity to failure is specific to the % of 1RM that is used? Why is this? These studies use 70, 75, 80, and 85% of 1RM and we know that the proximity to failure values vary at each VL value depending on the % of 1RM used (9). For example, VL20 will correspond to a different proximity to failure value at each 70, 75, 80, and 85% of 1RM.
To summarize, first, this data is suggesting what the optimal VL ranges are. As previously mentioned, the optimal VL range for strength and hypertrophy is ~0 – 25%. Second, this data is telling us, okay, since 0 – 25 VL is the optimal VL range, what proximity to failure does 0 – 25 VL correspond to at each % of 1RM. Consequently, we can determine what RPE we should be training at depending on the % of 1RM that is used and/or the number of repetitions that are performed in order to keep us in that 0 – 25 VL range.
How To Determine the Optimal Proximity to Failure?
Now that we’ve established that the optimal proximity to failure varies depending on the % of 1RM that is on the bar, how do we determine the optimal proximity to failure at each % of 1RM? First, I’ve created Table 2 to help conceptualize the optimal proximity to failure based on the % of 1RM on the bar.
Table 2. Relationship between 0, 10, 20, 40% Velocity Loss and RIR/RPE at 70, 75, 80, 85% of 1RM
The left most column titled “Group” provides the 4 VL groups used in the study by Pareja-Blanco et al. (2020): VL0, VL10, VL20, and VL40. Under each % of 1RM used in this study (70, 75, 80, and 85% of 1RM) I recorded the number of reps that were performed by each group.
Next, based on the smith machine squat data from Rodriguez-Rosell et al. (2019) I recorded the average number of reps performed at each % of 1RM. Specifically, in this study they reported an average of 23.4, 16.2, 9.6, and 6.0 reps at 50, 60, 70, and 80% of 1RM, respectively. In order to determine the average number of reps at 75 and 85% of 1RM (the percentages of 1RM used in the Pareja-Blanco et al. (2020) study) I ran a second order polynomial with % of 1RM on the x-axis and reps performed on the y-axis; providing me with a near perfect R2 value of 0.9984. Most importantly, it provided me with an average of 7.7 and 4.5 reps at 75 and 85% of 1RM, respectively.
Next, for each VL group I was able to determine the average RIR and RPE at each % of 1RM. For example, at 85% of 1RM, the average number of reps that can be performed is 4.5, and the average number of reps that were performed by the VL20 group was 2.3. Therefore, 4.5 subtract 2.3 is equal to 2.2 RIR, which equates to 7.8 RPE. To be clear, this table definitely isn’t perfect, and the RPE values within each group definitely varied greatly. Despite this, it provides a good average, upon which we can start to provide some training recommendations. For example, based on this data if we look at VL10 and VL20, the optimal RPE ranges from 70 – 85% of 1RM are 4.1 – 7.8, or if we round this to nice whole easy numbers about 4 to 8 RPE. You may ask, “How accurate is this?” Interestingly, if we look at VL40, the average RPE at 70, 75, 80, and 85% of 1RM is 8.0, 9.3, 9.3, and 9.8, respectively. In Pareja-Blanco et al. (2017), 56% of the total sets in the VL40 group were performed to failure, so I would argue that this average is pretty accurate and that perhaps the RPEs were even slightly higher than this.
I’ve combined data from these velocity loss studies numerous times in various different fashions in order to conceptualize the optimal proximity to failure at each % of 1RM. Interestingly, yet unsurprisingly, I always wind up with very similar values. For example at 70% of 1RM, I always wind up with about a 5 RPE for VL25. If we look at this table, at 70% of 1RM, I came up with about a 4.6 RPE for VL20. In other words, at 70% of 1RM, VL25 would be right about at a 5 RPE. Just a side note, but a cool side note nonetheless.
Integrating the Velocity Loss Data into the Proximity to Failure Data
Now that we’ve established that the optimal proximity to failure within 0 – 25 VL varies depending on the % of 1RM that is used, what is the minimum proximity to failure that we should be training at and what is the maximum proximity to failure that we should be training at? For example, since VL10 corresponds to ~4 RPE at 70% of 1RM, should we be doing sets with 70% of 1RM at ~4 RPE? In short, the answer is: no. I think you should be training at a minimum of ~5 RPE just to ensure that you are training close enough to failure in order to optimize strength and hypertrophy outcomes.
But first, why must we integrate the VL literature into the literature directly looking at different proximities to failure, and why should we not integrate the proximity to failure literature into the VL literature? There are two studies demonstrating significantly greater 1RM strength adaptations for high proximity to failure training over low proximity to failure training (10, 11). However, there are no studies demonstrating significantly greater 1RM strength adaptations between groups within the VL literature that looked at the squat (or smith machine squat) or the bench press (or weight stack bench press) aside from the pullup study which only compared VL25 to VL50.
Finally, a third study worth mentioning is Helms et al. (2018) in order to revisit our brief statistics lesson and see how despite no significant difference in 1RM between groups, there was still a moderate effect size (0.50) for the group that trained at a significantly higher RPE. If we were to solely look at the 1RM % changes from Helms et al. (2018) and compare those to the % changes from the VL literature, someone might falsely think that the magnitude of difference between groups was greater in the VL literature. For example, the 1RM % changes in the squat were 11.8 in the high RPE group and 10.0 in the low RPE group in the Helms et al. (2018) study. However, if you recall from our brief statistics lesson earlier, the effect sizes reported in this study are higher than the effect sizes reported in all of the VL literature, with the exception of bench press in the Helms et al. (2018) study (effect size: 0.28) compared to smith machine squat in the Pareja-Blanco et al. (2017) study (effect size: 0.34).
Conclusively, for this reason, we must first use the optimal RPE range of ~5 – 10 first, and the optimal VL range of ~0 – 25 second. Okay, well how did I determine the optimal RPE range? The high RPE group in the Helms et al. (2018) study trained at ~7 – 9 RPE for the majority of the study; however, RPE values are overestimated, yet improve as the number of reps in the set decreases (therefore at higher percentages of 1RM) and as the proximity to failure increases. However, the study that noted this finding performed more reps on average at a lower % of 1RM than what the subjects trained at in the Helms et al. (2018) study; therefore, for this reason it is difficult to make a direct comparison. However, I would estimate that these subjects probably overestimated their RPE values by ~2; however, some probably underestimated their RPE values and may have even been training close to a 10 RPE. Therefore, I think that the optimal RPE range is ~5 – 10 RPE. Now, that we’ve established the optimal RPE range, we can input the optimal VL range of ~0 – 25% into the optimal RPE range and answer two key questions: 1) what is the optimal RPE range for the % of 1RM that I am using? And 2) what is the optimal RPE range for the number of reps that I am performing?
General Training Recommendations
I have used the data from Rodriguez-Rosell et al. (2019) in order to provide some general training recommendations and answer the two aforementioned questions. Importantly, this data literally aligns perfectly with table 2, further supporting that the approximate RPE values that I determined for each VL group in the Pareja-Blanco et al. (2020) study and Pareja-Blanco et al. (2017) study were probably pretty accurate. First, I’ll begin with the optimal RPE ranges for each % of 1RM. Second, I’ll address the optimal RPE ranges for each rep performed. However, for training purposes, please remember to use your own individualized first rep velocity table and last rep velocity table for accurate RPE values and percentages of 1RM, as addressed in my article titled SYSTEMATICALLY INDIVIDUALIZING LOAD PRESCRIPTION: FORMULATING AND APPLYING FIRST AND LAST REP VELOCITY TABLES.
Table 3 illustrates the optimal RPE range at each % of 1RM in order to maintain 0 – 25% VL. The recommendation starts at 70% of 1RM, because this is maxing out the bottom end of the optimal RPE range (5 RPE) and maxing out the top end of the optimal VL range (25%). For each 2.5% increase in 1RM the top end of the optimal RPE range increases by 0.5 RPE. Therefore, the optimal RPE range for 72.5% of 1RM is 5 – 5.5 RPE, 75% of 1RM is 5 – 6 RPE… 95 – 100% of 1RM is 5 – 10 RPE.
Table 3. Optimal RPE Range at Each % of 1RM to Maintain 0 – 25% Velocity Loss
Table 4 illustrates the optimal RPE range for each rep number to maintain 0 – 25% VL. The recommendation starts at 6 reps, because this is maxing out the bottom end of the optimal RPE range (5 RPE) and maxing out the top end of the optimal VL range (25%). For each decrease in reps the top end of the optimal RPE range increases by 1 RPE. Therefore, the optimal RPE range for 5 reps is 5 – 6 RPE, 4 reps is 5 – 7 RPE… 1 rep is 5 – 10 RPE.
Table 4. Optimal RPE Range for Each Rep Number to Maintain 0 – 25% Velocity Loss
If the primary goal of the session is strength, you’ll want a combination of high RPE (~7 – 10) and low VL (~0 – 20). Why high RPE for strength? There are two studies demonstrating significantly greater 1RM strength outcomes for higher RPE training, in which subjects trained at ~7 – 9 RPE for the majority of the study in both studies (10, 11). Why low VL for strength? There is one study demonstrating a moderate effect size in 1RM strength outcomes for VL20 over VL40, and below 20% VL is not inferior for strength, its simply that those groups probably didn’t perform enough volume to have as robust increases in 1RM strength as VL20 (3, 6).
If the primary goal of the session is hypertrophy (for a strength athlete) you want a combination of low RPE (~5 – 6) and high VL (~20 – 25). Why high VL for hypertrophy? VL20 resulted in a significant increase in hypertrophy relative to baseline; however, VL10 did not (6). Also, there was no significant difference in hypertrophy reported between VL20 and VL40 (3, 6). Therefore, it appears that there is no added benefit for training at higher VL thresholds. In addition, higher VL thresholds will promote negative neuromuscular adaptations for strength athletes (3). Why low RPE for hypertrophy? First, there was no significant difference nor any meaningful effect sizes reported by Helms et al. (2018) in hypertrophy between the high and low RPE groups; however high RPE was superior for strength (12). Second, in order to maintain that 20 – 25% VL, but still get in sufficient volume via higher reps (~5+) while maintaining an RPE of 5 or greater, you’re really only left with a few options: 6 reps @ 5 RPE (~25 VL), 5 reps @ 5 RPE (~20 VL), and 5 reps @ 6 RPE (~25 VL). If you perform less than 5 reps, it will be difficult to get in sufficient volume; however, if you perform more than 6 reps, you will no longer be within both the optimal RPE range and optimal VL range. For example, 7 reps at a 5 RPE is ~30 VL (exceeding the top end optimal VL range). Similarly, 7 reps at 25 VL is at a ~4 RPE (exceeding the bottom end of the optimal RPE range).
Table 5. RPE Range, Velocity Loss Range, and Repetitions Range for Training Goal
The velocity loss literature suggests that there is not one magical optimal proximity to failure, but rather that that the optimal proximity to failure is contingent upon the % of 1RM on the bar and the number of reps that are performed within the set. If you enjoyed this article and are interested in how you can individually optimize your own proximity to failure specifically for powerlifting, please check out my Bonus Seminar. Enjoy!
- The majority of training for strength athletes should be performed at or below ~25 VL in order to preserve favorable neuromuscular adaptations
- The majority of training for strength athletes should be performed in the 1 – 6 rep range, 5 – 10 RPE range, and 0 – 25 VL for optimal strength and hypertrophy outcomes
- Athletes can use the optimal RPE range tables provided as a general guideline in order to maintain 0 – 25 VL depending on the percentage of 1RM on the bar and number of reps performed within the set
Davies, T, Orr, R, Halaki, M, and Hackett, D. Effect of training leading to repetition failure on muscular strength: a systematic review and meta-analysis. Sports Medicine 46(4): 487 – 502, 2016.
Pareja-Blanco, F, Sanchez-Medina, L, Suarez-Arrones, L, and Gonzalez-Badillo, JJ. Effects of velocity loss during resistance training on performance in professional soccer players. International Journal of Sports Physiology and Performance 12(4): 512 – 519, 2016.
Pareja-Blanco, F, Rodriguez-Rosell, D, Sanchez-Medina, L, Sanchis-Moysi, J, Dorado, C, Mora-Custodio, R, Yanez-Garcia, JM, Morales-Alamo, D, Perez-Suarez, I, Calbet, JAL, and Gonzalez-Badillo, JJ. Effects of velocity loss during resistance training on athletic performance, strength gains and muscle adaptations. Scandinavian Journal of Medicine & Science in Sports 27(7): 724 – 735, 2017.
Galiano, C, Pareja-Blanco, F, Hidalgo de Mora, J, and Villarreal, ES. Low-velocity loss induces similar strength gains to moderate-velocity loss during resistance training. The Journal of Strength and Conditioning Research, [Epub ahead of print], 2020.
Rodriguez-Rosell, D, Yanez-Garcia, JM, Mora-Custodio, R, Pareja-Blanco, F, Ravelo-Garcia, AG, Ribas-Serna, J, and Gonzalez-Badillo. Velocity-based resistance training: impact of velocity loss in the set on neuromuscular performance and hormonal response. Applied Physiology, Nutrition, and Metabolism, [Epub ahead of print], 2020.
Pareja-Blanco, F, Alcazar, J, Sanchez-Valdepenas, J, Cornejo-Daza, PJ, Piqueras-Sanchiz, F, Mora-Vela, R, Sanchez-Moreno, M, Bachero-Mena, B, Ortega-Becerra, M, and Alegre, LM. Velocity loss as a critical variable determining the adaptations to strength training. Medicine and Science in Sports and Exercise 52(8), 1752 – 1762, 2020.
Rodiles-Guerrero, L, Pareja-Blanco, F, and Leon-Prados, JA. Effect of velocity loss on strength performance in bench press using a weight stack machine. International Journal of Sports Medicine, [Epub ahead of print], 2020.
Sanchez-Moreno, M, Cornejo-Daza, PJ, Gonzalez-Badillo, JJ, and Pareja-Blanco, F. Effects of velocity loss during body mass prone-grip pull-up training on strength and endurance performance. The Journal of Strength and Conditioning Research 34(4), 911 – 917, 2020.
Rodriguez-Rosell, D, Yanez-Garcia, JM, Sanchez-Medina, L, Mora-Custodio, R, and Gonzalez-Badillo, JJ. Relationship between velocity loss and repetitions in reserve in the bench press and back squat exercises. The Journal of Strength and Conditioning Research, [Epub ahead of print], 2019.
Shattock, K, and Tee, JC. Autoregulation in resistance training: a comparison of subjective versus objective methods. The Journal of Strength and Conditioning Research [Epub ahead of print], 2020.
Graham, T, and Cleather, DJ. Autoregulation by “repetitions in reserve” leads to greater improvements in strength over a 12-week training program than fixed loading. The Journal of Strength and Conditioning Research, [Epub ahead of print], 2019.
Helms, ER, Byrnes, RK, Cooke, DM, Haischer, MH, Carzoli, JP, Johnson, TK, Cross, MR, Cronin, JB, Storey, AG, and Zourdos, MC. RPE vs percentage 1RM loading in periodized programs matched for sets and repetitions. Frontiers in Physiology 9: 247, 2018.
Landyn Hickmott, MS, CSCS
Powerlifting Athlete, Coach, & Researcher
Landyn Hickmott is an emerging researcher specializing in individualization, autoregulation, and
velocity-based training as it pertains to the sport of powerlifting. Formerly, he was a semiprofessional hockey player in the Western Hockey League. Currently, he is a nationally-qualified
powerlifter in the Canadian Powerlifting Union and a graduate researcher in the Muscle
Physiology Lab at Florida Atlantic University. He will be graduating with a thesis-based
Master’s of Science in August and will be pursuing a PhD at the University of Saskatchewan in
January. Landyn created Individualized Adaptive Powerlifting Programming; a holistic training
system that conceptually integrates RPE and VBT in order to develop an individualized and
adaptive model that is optimally customized for each powerlifting athlete.
Connect with Landyn here!