Building Metric: A computer vision app for VBT

I have been obsessed with velocity based training for as long as I have been coaching.

We had a linear positional transducer (tethered string device, LPT for short) in our gym very early on and would have our athletes compete on relative power output. Whenever an athlete started to hit a plateau for weight, we would shift focus and chase watts relative to bodyweight as a way to create a new training goal, making sessions competitive and motivating.

We knew we were on to something, because every time we brought out the LPT device for a few weeks, athletes started telling us about how much higher they were jumping. Chasing power was giving their lifting intent a boost and that translated into improvements in rate of force development and athletic power output.

Unfortunately, we only had one device and with over 100 athletes we couldn't warrant bringing it out every session. The slow software, bluetooth issues and cost of getting more devices meant we had to treat velocity tracking as a novelty and not a staple of our training.

Despite these limitations, our great results showed just how valuable even this simple application of VBT could be. It makes so much sense that we should put as much attention into the velocity and intent of our lifting as we do into the load, reps, and sets we complete.

You already do velocity based training - you just don't call it that

Every person who has been lifting weights for even a short period of time uses velocity to regulate their training. Whether it is conscious or subconscious, lifters are constantly analysing their sets through the lens of how fast the bar moves.

After, or even during the set, lifters perform system checks, analysing the feedback from their bodies and brains to help answer the important question about how challenging a weight is, trying to gauge how close the set was to failure, and how well we are performing relative to our recent history.

The language we associate with this effort almost always refers to the speed, or smoothness of the set. Effort, intent and exertion are all described by describing how fast the weights move.

"Those deadlifts flew up"

"I ripped that power clean"

"The weights are popping today"

"That 180kg was sloooow today"

Athletes and coaches have intuitively understood the connection between bar speed, effort levels and proximity to failure long before velocity tracking became a thing. When the weights are moving fast we have plenty left in the tank, so we load up the bar or smash out more reps, but when the weights are moving slow, we are closer to our one-rep max (1RM) or to the point of failure. Even a moderately experienced lifter or coach can judge this relationship between velocity and exertion with their coaches eye and a learned feel for the lift.

It is therefore the logical progression that we would take this subjective assessment of bar speed and make it objective. Velocity based training (VBT) is simply using technology to accurately measure your movement velocity during training, fixing a number to the quality of a set, giving us additional data points to precisely assess training quality.

In the same way that weight plates have their weights written on the side so we can add up the correct load for our training, tracking movement speed, power output, tempo, or range of motion are just extra objective labels that can aid in monitoring progress and inform better training decisions both in-session and in long term periodisation.

Velocity based training isn't about training fast

When first learning about velocity based training it is common for lifters and coaches to be misdirected in what it means to use velocity in a training plan. I think this confusion largely comes from VBT being an inappropriate name.

The name velocity based training implies training at high speeds. Lifting light loads for explosive high speed repetitions - jump squats, bench throws, Olympic lifts, things of that nature. And while yes you can quantify performance on dynamic effort exercises with VBT, it's not actually the best way to utilise velocity tracking.

Train as normal, velocity just adds context

The rules of smart training and program design don't cease to exist when you start tracking velocity, to get strong you have to lift heavy weights, and those weights are going to move slowly. Whether you use velocity tracking or not, strength programming should centre around progressive overload and the pursuit of strength gains - largely through increasing the number of plates on the bar. What velocity tracking does is provide an objective insight into an athlete's readiness to train and the quality of their intent during that training, potentially priceless data that can further optimise training and outcomes in real-time.

Velocity tracking does not stop at simply bar speed either, we can use the precise tracking of lifts to gain insight into range of motion, power output, tempo, time under tension, bar path, fatigue accumulation, 1RM estimations, performance profiles and much more. These data points not only support better strength and power training but also refining movement proficiency, hypertrophy and bodybuilding, tapering for athletic competition, rehabilitation, and athletic performance testing.

Velocity tracking allows coaches and athletes to do three fundamental things with their training:

• Motivate lifters to move all sets with maximum intent. Increasing rate of force development and motor unit recruitment.
• Auto-regulate training based on real-time assessment of your fatigue and readiness status to help lifters find the optimal training load or volume for a given session.
• Track progress over time in multiple dimensions, moving away from the single dimension of load on the bar into a world where every repetition becomes a data point that allows you to assess progress towards training goals.

Motivating lifters to move with intent

Intent to move is an interesting physiological phenomenon well known in strength and conditioning. We know that the intention to move as fast as possible impacts nervous system adaptation more than the actual speed the athlete moves.

As an example, imagine your are in a strength training block, lifting loads well above 90% of your estimated 1RM for sets of two or three repetitions. These working sets are never going to be fast in absolute terms, for this athlete these heavy back squats or bench presses might have a mean velocity of say 0.3m/s.

Understanding the intent to move phenomenon, you should still try and lift that heavy weight as fast as you possible can despite knowing the weight will prevent you from actually achieving a fast velocity. Even if your absolute best effort only reaches say 0.32m/s, as long as that 0.32m/s is indicative of your very best intended velocity for that given weight and exercise, you will unlock greater strength and power adaptations than had you just used the bare minimum intent and completed each rep at only 0.28m/s.

The slight difference in bar speed is of little relevance, the fact that the set with higher intent to move led to 100% of your motor units being activated instead of just 85-90% will have a huge impact on improved neural output and lead to significantly greater gains.

Before velocity based training enabled precise tracking of bar speed we had to estimate how much intent an athlete was putting into their lifts, giving subjective feedback, and motivating them with loud music and ra-ra coaching. Velocity tracking allows athletes and coaches to quantify the intent to move for every set with objective precision. We can now compare an athletes effort today with last week and immediately flag when they are lifting with intent or just going through the motions or are under a fatigue cloud and might need to autoregulate their training.

Autoregulation of training

Making progress in the gym is rarely a purely linear activity. It's pretty tough to show up every week and do one more rep or add one more kilo to the bar after those first six months of newbie gains start to slow down.

Some days we destroy the gym, lifting 120kg for five reps like it is nothing; the bar is flying up, RPE-nothing. On the less good days, that same 120kg feels like 220kg. We grind out a shaky, ugly triple, and it feels like we will never hit a new personal record again, RPE-11.

This is due to our readiness to train, a concept that takes into account weekly, daily and even hourly fluctuations in strength and power levels that play a huge role in how well we perform physically. Stress, sleep, nutrition, hydration, residual fatigue, mental state, pain, and injury can all have an impact on this readiness to train, and some research suggests it can vary as much as 18% day to day (Jovanović M, Flanagan E. 2014) although more commonly 5-10% fluctuations (Zourdo, M, et al. 2016) are seen.

Our improved understanding of fluctuating readiness is a huge part of why percentage based training is falling out of favour and is being replaced with more opportunistic periodisation models like using velocity data or reps in reserve (RIR) to adjust training plans in real-time.

Reps in reserve (RIR) training is a method for measuring your readiness to train by calculating your proximity to failure for the set. RIR-1 is one rep shy of failure, the equivalent of RPE-9. Observing RIR (or RPE - rating of perceived exertion) allows athletes and coaches to regulate training loads and volumes, making micro adjustments to the planned session based on how fatigued they are. This helps optimise the training stress and give athletes more time to recover on the days when they need it.

Unfortunately RIR/RPE based measures are inherently subjective. Humans can be horribly biased when it comes to self-assessment and can be susceptible to psychological anchoring, social pressure, and ego when determining how hard a set was. We are much more likely to round an RIR up (or RPE down), telling our coach and ourselves that the set was easier than it really was to avoid having them cut a training session short.

Coaches and athletes will improve at this over time, and they can fight this bias by applying a formula or pre-determined algorithm to the RIR data, taking the decision making out of their hands, turning regulation into autoregulation. Simply put, an algorithm takes human emotion out of the decision making process by applying a pre-defined "if this happens, then that occurs" model.

Unlike RIR, accurate velocity metrics are objective measurements and very hard to cheat. For example, let's say last week's best-rep velocity for my 150kg deadlift was 0.60m/s, and today it is 0.62m/s. This improved velocity could suggest three things: I have gotten stronger, I am lifting with more intent, or I am fully recovered from my last session. Whichever combination of the three variables it is, this is a positive result and a green light for today's session.

Alternatively, if today's velocities are significantly down compared to my recent history, say 0.52m/s, a 15% drop from last week, then maybe today is not the day to try for a new PR. Instead I should maybe just repeat the same weights as last session, or even cut a rep off each set for a tiny 1-2% deload.

It's not that we only use velocity and disregard all other training variables or stop using RIR/RPE, instead adding velocity into the data mix can further optimise our training, keeping the RPEs honest, and gaining a clearer picture of if we are tracking in the right direction.

Tracking progress over time

While the goal in strength sports is pretty self-explanatory (get stronger!), exclusively chasing more and more weight every session is a near-sighted approach. A "plate-hungry" approach to training rewards consistently pushing closer to failure, compromising technique in favour of load, and accumulating fatigue by grinding out extra reps that tax the nervous system more than they develop it. In fact, there is a growing body of evidence that this weight-at-all-costs approach is counterproductive to the strength adaptations we are chasing. You may lift a few extra kilograms this week but it will hurt your progress in the long run.

Velocity provides an extra dimension to scoring our progress in the gym. With regular velocity tracking every set recorded becomes a data point to help us measure if we are improving. With this data we can now measure progress in a number of ways:

• Lifting heavier loads
• Completing more reps
• Lifting the same weight but doing it faster than last week
• Greater power outputs on any load

These are basic measures of progress between sessions, but we can go another level up and use velocity tracking to score progress using the load velocity profile, a topic for another time.

The problem with tracking velocity

If velocity based training is so useful and versatile, why isn't it more widely adopted? Why didn't we spend the money and get more LPT units all those years ago?

As of writing, VBT is quite popular in the college training environment and at the professional level but it hasn't really filtered down to the wider training community. I think there are a few reasons for this:

• The notion that velocity tracking should be complex and sophisticated. Many VBT advocates portray the method as if it were a dark art. They reel off velocities, ratios, charts, and data in rapid fire, creating an intimidating knowledge hurdle for people to overcome in order to start using VBT.
• Expensive, cumbersome hardware. Legacy solutions require an additional piece of specialised hardware to be purchased, sometimes costing thousands of dollars. Price is an obvious barrier here, and while prices have certainly come down in recent years, equipping a medium sized private gym with multiple units is still not cheap. Beyond price, hardware comes with another set of logistical burdens: batteries to keep charged, swapping from bar to bar, connecting to the right tablet, and switching profiles all add to the friction of velocity tracking.
• Poor software experience. As inconvenient as hardware can be, having a poor software experience can be twice as frustrating. The data that can be gleaned from velocity tracking is incredibly valuable, but often times the effort and friction involved in accessing and interpreting the data significantly impedes the training and coaching flow. Unless you have a team of interns transposing the data into Excel it can be incredibly hard to action the velocity data in real time, and that is where the benefit is.

Making VBT accessible

This is why we are building Metric VBT.

Metric is the world's first fully automatic phone-based computer vision VBT solution.

No strings. No additional hardware, just point your camera, hit record and lift.

Metric completed it's Beta testing phase in April 2022 and is now available on the Apple iOS app store for iPhone and iPad. The app is free to download and you can record unlimited sets with a free Metric account. Accurate and easy to use velocity based training can be done for $0.

Metric is now available on the iOS App Store, you can download it for free at this link!

With Metric we are completely rethinking how velocity based training is done in the weight-room. Our goal is to address the pain points with VBT by making it effortlessly usable, providing rich contextual data, seamless video integration, and actionable real-time training improvements. We have gone back to the absolute foundations on everything velocity to reimagine how athletes train with velocity and how coaches program. From how you measure velocity, the metrics that matter, scoring and profiling progress, through to app design, training flow, and what data you see and when.

We aren't making a slightly better option, we are completely changing the VBT game.

This is the most important section of any article on this website, but it's also one of the shortest. When you spend two years fixated on a problem your thoughts tend to become incredibly focused. We cannot wait to share MetricVBT with the world.

How we made MetricVBT

But the Metric vision hasn't always been this clear. wWe need to go back to the very beginning to understand how we got here. It all started when my brother Davey started coming to Core Advantage to workout.

Davey is proper smart. A computer scientist and electrical engineer by training, Davey is the kind of person who 3D prints a puzzle to hide your birthday money in.

He has been working at Core Advantage since 2016 and has built so many incredible things to streamline our high performance and rehabilitation coaching services, including simple admin integrations, the Core Advantage Interval App, and our custom timing gates system and athlete testing database.

Davey is the behind-the-scenes hero of our business, quietly going about building these brilliant sports science solutions with a really good lateral approach to long standing fitness industry problems.

In 2016, not long after he returned from six months studying abroad in Sweden, we started training together. Davey would come to the gym after hours, we would lift some weights, talk nerdy stuff and hang out.

I was lifting with a tethered VBT device and after some back and forth on the benefits of velocity based training Davey was a bit disappointed that we were stuck using a string to measure velocity.

Then he says "What if we used SONAR to measure bar speed?".

"You mean like a bat?"

So he got to work, spending a few days a week developing an accurate wireless solution to track bar speed using SONAR. We made a pendulum style cardboard box with a sensor facing the ground and hung it off the barbell.

In the few tests we ran, it was actually pretty accurate. We got consistently within 85% of an LPT on day one. Not valid, but reliable enough to be an interesting idea.

The tricky thing with research and development is that it is expensive.

Going from an interesting idea that has potential to creating something that will really work is incredibly costly in both hours and dollars. Then to take the next step and go from something that works to an actual product we could sell and ship is twice as hard and expensive again. So as excited as we were with the prototype and potential of our SONAR VBT, Core Advantage in 2016-2017 was still very much a gym first and foremost, we weren't thinking about becoming a technology company.

So the SONAR (we were calling it the bar-bat) went from being an interesting idea, to a dormant project, to being completely shelved.

As interesting as this solution was, we started to realise that the problems with VBT couldn't be solved by building different hardware. The problem was the fact that you needed any hardware in the first place. Having to buy, carry, set up, charge and care for an extra physical device was the key limitation for most people starting with VBT.

The business model and work required to find out whether people would like a SONAR VBT device wasn't worth the gamble, so we returned our focus to building in-house solutions: the testing software, our timing gates, and the Core Advantage Interval App.

But Davey and I continued training together a few times a week, talking technology and getting stronger, all the while complaining about having to use a string whenever we trained with velocity.

Then GymAware launched the Flex unit in October 2019.

Initially we were angry, they had beaten us to the punch with an accurate stringless velocity tracker! But we quickly realised that their idea of an array of optical sensors around the bar was a much smarter idea than our pendulum SONAR sensor. Plus Kinetic performance build beautiful hardware.

Our anger and envy quickly turned to excitement though, because it got us back to thinking about how to solve the velocity tracking problem again.

One day between bench sets Davey says: "We could probably track a barbell with your smartphone camera using computer vision?"

"You mean like Tesla auto-pilot?"

Computer vision in general, and especially phone-based computer vision had come a long way since we first explored velocity tracking in 2017, and augmented reality continues to improve at an accelerating pace, but we probably weren't prepared for how hard it would be to achieve an accurate technology!

Nevertheless we knew that getting accurate velocity data without a device, all while collecting video footage on your phone could revolutionise and democratise velocity based training.

So we got to work, building and validating our computer vision solution over about 18 months. In September 2021 we released the first teaser of Metric VBT, the worlds first automatic velocity based training computer vision app.

We are well on our way in solving the computer vision problem, with the private beta right around the corner, but there is a secondary problem that Metric aims to solve which is ease of use. Velocity tracking should be both easy to implement and beneficial to the training process, something not all existing technologies deliver on. Over the coming months and years we plan to solve that problem building an affordable and friction free velocity tracking solution.

Learn more about Metric

Metric can be downloaded for iPhone via the App Store and directly from this link.

UPDATE: Since publishing this blog we have released the first internal validation of MetricVBT, check it out here.