As we’ve collected data through our force plate strength assessment, trainees commonly ask exactly how those physical qualities show up in their sport skill (hitting or pitching). Correlating metrics from each test yields moderate positive correlation with skill outcomes, but we wanted to take this a step further.
To attempt to answer this question, High Performance R&D Intern, Anthony Osnacz used our force plate data to investigate how the physical qualities measured in our jump and strength assessments show up on the mound or in the batter’s box.
The current model we use to predict athlete pitch velocity from our force plate assessments takes physical qualities in a vacuum—absent any skill, intent, readiness, or any of the many other factors important for a pitcher. Anthony’s models are not intended to tell us, “If you have x force plate metrics, you will throw y velocity.”
Rather, they say, “If you have x force plate metrics, we would expect you to throw y velocity, with all other variables equal.” This allows trainers to quickly determine whether an athlete’s actual mound results are out performing or underperforming their predicted fastball velocities, therefore allowing us to quickly identify and communicate lowest hanging fruit for a pitcher’s training.
Athletes A and B (velocities shown in table below) have a similar predicted fastball velocity from their force plate profiles, but their actual motion capture tests are 16mph apart. This helps demonstrate how two similar force plate tests can result in different training recommendations.
|Athlete A||Athlete B|
|Predicted Fastball Velocity (mph)||87.9||87.2|
|Actual Motion Capture Velocity (mph)||95.8||79.7|
|Predicted Velo – Actual Velo (mph)||-7.9||+7.5|
The typical interpretation for Athlete A’s mound velocity outperforming his predicted velocity was that his throwing skill was outperforming his general physical qualities, and with that in mind his lowest hanging fruit this offseason was improving the physical areas he lacked the most. This meant allocating more of his training economy to building strength and power in the weight room and less training economy to specific throwing work.
Athlete B is in essentially the opposite situation, with his mocap velocity underperforming his predicted velocity. Typically, an athlete in this situation needs a focus on the skill side, so this offseason he had throwing work emphasized with reduced volume in the weight room reflected in his training economy balance.
Inception of predicting velocity
The idea for predicting velo based on force plate metrics was inspired by an investigation by Alex Caravan back in May, in which he used ordinary least squares regression to create a model to predict throwing velocity off of force plate metrics. Benefiting from a larger sample size we decided to re-run the analysis.
We split the data 75/25 into training and test data frames respectively. We used linear regression on the training data to find a model and then tested that model on the test data frame. The model had a 2.78 mean absolute error, which means on average the model was +/- 2.78mph off of an athlete’s actual average MoCap Velocity. This showed us that the model had some real potential.
The metrics that are weighted the heaviest in the model are Squat Jump Peak Power in Watts, RSI-Modified in m/s recorded during the Counter Movement Jump test, and Net Peak Force in Newtons from the Isometric Mid-Thigh Pull. Squat jump peak power is simply the peak amount of power an athlete generates during their squat jump. In our assessment process an athlete performs 4 trials and then the peak power from each trial is taken and the mean is found and what is used in our model.
RSI-Modified is a power metric that is found by dividing an athlete’s countermovement jump height in meters by their time to take off in seconds. Again, we perform 4 counter movement jumps and find the RSI-Modified for each trial and then take the average of them and use that in our model. Net Peak Force is found by subtracting an athlete’s body weight in Newtons from the Peak Force generated during their best Isometric Mid-Thigh Pull trial.
To illustrate just how strong the relationship is observe the graphic below and check out this excellent Twitter thread from Dan Adams. Stronger and more powerful athletes tend to throw harder.
And to further demonstrate the strength of the relationship let’s filter the data to just athletes who throw 88+ mph. As you can see below, not a single 93+ mph thrower possesses less than above average Squat Jump Peak Power.
As mentioned above, predicted velocity based purely off of force plate metrics can give us more insight into which qualities we should strive to improve in the athlete’s training. In addition, it can be used as an approximation of an athlete’s “engine”. If an athlete is throwing harder than their predicted velocity, you could infer that their lowest-hanging fruit might be on the strength and conditioning side of things, whereas the lowest-hanging fruit for someone who is throwing below their predicted velocity may be on the skill side of things.
Current limitations of the model include not being able to account for how skillful an athlete is at throwing (though the point is to exclude that), not accounting for anatomical differences such as limb length, and not having any information on an athlete’s upper body strength or power.
In the future we plan to measure athletes’ limbs and use that data to improve the model’s performance. The longer a lever is, the greater the force on a load will be. So an athlete who has longer limbs, all else being equal, will generate more force when throwing, thus resulting in a higher velocity. This is something that is important to account for when trying to estimate an athlete’s propensity to throw X mph.
Another way in which we plan to improve the model is potentially through the use of Force Hooks. If validated, we can gather data on an athlete’s absolute upper body strength through an isometric bench press. This could give us a better picture of an athlete’s physical capabilities, which, in theory, would make for a better model.