Examining Movement Strength Ratios: How Data Science is Changing the Landscape of Strength and Conditioning

Phase 31246.jpg

Successful strength programming requires the balance of scientific principles and practical real-world methods. While scientific inquiry drives our understanding of training and injury prevention, it is the coach’s ability to implement that science that determines how effective a training program is. The balance between the two is what allows coaches to make the greatest impact possible. Often, the use of practical shortcuts is necessary to make training effective in constrained environments. The demand for strength coaches to be economical with time and energy makes the use of simple heuristics necessary. Volt was built utilizing many of these same strategies, but with the integration of our performance training AI, Cortex, Volt can help bridge the gap between science and practice in remarkably new ways.


Balancing Science and Practice

Having an understanding of an athlete’s one-repetition maximum (1RM) provides information on their upper-limit of strength in a movement. But there are so many movements that would need to be tested in order to get a full picture of the athlete’s actual strength levels, so the practice of using a "parent" movement to determine another (or several) movement's 1RM became a common strategy. This allows coaches to focus more energy and instruction on more consequential movements while still deriving a suitable "working weight" for less consequential movements.

An easy example is to use the back squat as the primary influencer to what weight an athlete should use for a lunge or split squat. Comparatively, the back squat is a better indicator of pure strength, uses more absolute load, and is traditionally more common in the bulk of a program— and thus requires more intense coaching and instruction to get right. The lunge and split squat are typically utilized as accessory movements, for they help support progress in the squat but do not impart as much systemic stress. When a coach uses the back squat 1RM to inform loading for the lunge and split squat, there is a gain in efficiency but a loss of specificity.

Accessory movements become completely reliant on the result of the parent movement, and lack a rate of progression that may be unique themselves. Accessory movements can be tested up to a specific repetition max but it typically is not advised, and is unnecessary. It's more important to simply get a starting weight for those movements so that they can be performed and progressed as the athlete demonstrates more proficiency in the squat to dictate the rate of overload. It's an efficient method but there is an opportunity to be more effective.


Understanding Movement Ratios using Artificial Intelligence

Volt was originally built with this same logic in place, but on a much broader scale. Each loaded movement in the Volt library is linked to other "related" movements, which provided a cohesive network of movements that share threads of similarity. This network of related movements gives Volt an incredible advantage when analyzing the "degree of relation" of each movement in our training system. Being a product built from a lineage of strength and conditioning knowledge, our initial three parent movements were the Barbell Back Squat, Barbell Bench Press, and Barbell Hang Clean. We've been developing and modifying a proprietary algorithm for the last six years, building an intricate network of proprietary ratios that connect our library of movements to these parent lifts. We've been slowly improving each year as more and more athletes train with Volt, providing larger datasets to determine the effectiveness of the loading parameters.

Email 1_ Welcome - Cortex feature (5).jpg

But in 2018, our newly integrated AI system, Cortex, was released and has allowed us to utilize auto-regulation in our interface. We now capture a user’s RPE (Rate of Perceived Exertion) for each loaded movement, providing an easy metric to compare the perceived difficulty of a movement against our prescribed recommendation. Understanding the user's feedback created an amazing opportunity. With Cortex, we could analyze the discrepancy between the "intended" difficulty of movements and the "perceived" difficulty. This moved us to design experiments to determine the accuracy and utility of our proprietary algorithm with a more detailed understanding of the user’s perception. With Cortex, the speed at which we can analyze data was magnified to the point of the largest "living training lab" available. With over a million athletes training on Volt, we are acquiring a massive dataset that informs us on the level of "strength relatedness" across all of Volt's loaded movements.


Our Findings

The results were surprising and validating.


For every movement we related to the Barbell Back Squat, our prescribed movement load and intended RPE for the first set was within 1.3% ± 1.3% of what users actually recorded. Only when we predicted the Barbell Deadlift from the Barbell Back Squat did we see our estimates deviate by a greater margin (about 6%). These results suggest that our derived ratios between lower body movements accurately reflects user reported performance.


For any movement related to the Barbell Bench Press, the average difference in estimated load at a prescribed RPE was incredibly small, 0.9% ±0.4%. This suggests an even closer ratio between our suggested level of intensity compared to a user's reported performance.

Initial loading for movements related to the Barbell Hang Clean carried an average difference of 1.5% ± 1.2%, with only one movement demonstrating a ratio greater than 2.5% (RDL to Barbell Hang Clean was on average 5.9%). Similar to our results to lower body movements, our movement to movement ratios for the first set aligned well with a user’s reported performance.

In conclusion, when we prescribe a set for a user by knowing nothing other than their performance on a different movement, our prescription tends to be incredibly accurate in predicting the subjective difficulty of the set.

This was simply a snapshot at one point in time, and many more questions need to be asked as we continue to refine our training algorithms. With our commonly used movement ratios showcasing a reliable accuracy, we're excited to run more analyses. Answering these questions brings the science of training to every user.


Future Analysis

Volt will continue to deliver a cutting-edge training experience and maintain our hard work behind the scenes analyzing the results. Volt is helping to uncover new insights and understandings about training that can help coaches advance beyond simple heuristics and upgrade their training toolkit. As more and more athletes leverage Volt's training, the number of potential discoveries that would once have been hidden are growing daily. We'll be presenting our findings from these analyses at the 2019 NSCA National Conference in Washington D.C. and look forward to continuing to expand our ability to leverage AI into the world of strength and conditioning.


Join over 1 million coaches and athletes using Volt's AI-powered training system. For more information, click here.

Jace Derwin, CSCS, RSCC, is the Head of Performance Training at Volt Athletics and is one of the regular contributors to the Volt blog. Jace manages Volt program design, content development, and educational resources for schools, clubs, and organizations. Jace is a Certified Strength and Conditioning Specialist® (CSCS®) and holds a bachelor’s degree in Exercise Science from Seattle Pacific University. Follow Jace on Twitter @VoltCoachJace.