01

November



Share

Share

Biomechanical Equivalence: Projecting batting statistics using biomechanics data

At Motus, we specialize in biomechanical analysis and data science. Over the last decade, we have contracted our gold-standard motion capture lab services to capture swings and mobility exams on thousands of baseball players. While we use this data to help inform training decisions, we have stealthily contracted more prestigious services to professional organizations: player projections powered by biomechanics data. We call this projection system “Biomechanical Equivalence”, and it gets more predictive with each dataset we collect.

During a biomechanics exam, we use our optical motion capture system (16 MAC Raptor E Cameras and 2 Bertec Force Plates) to capture movement data. The cameras operate at 480 Hz and can function in direct sunlight (allowing us to capture data in a more natural environment).

A typical biomechanics exam at Motus consists of a movement screen and a swing analysis. The movement screen is a battery of 25 active range-of-motion movements of the shoulders, spine, hips, and ankles. The swing exam consists of a live batting-practice session of 20 swings, where we collect full-body biomechanics, weight distribution, bat dynamics, and batted ball dynamics. The result of the total exam is over 500 discrete measures, as well as “tens-of-thousands” of raw time-series kinematics data points.

Sample biomechanics data from range-of-motion testing and swing analysis.

Using these data points from every player in the Motus database as inputs, we then segment all subject data into competition level (i.e. Rookie, Low A, A, High A, AA, AAA, MLB). Differences by level certainly exist for certain discrete measures. Within groups, game-statistic correlations also emerge.

However, we can take this a step further by using Bayesian modeling. We begin inputting all of the segmented data points and batting statistics and produce multi-variate equations to project a player from one level of competition to the next. The resulting set of equations allows us to analyze a player at the rookie level, and project his performance at the MLB level given his current biomechanics. This projection system is “Biomechanical Equivalence”, and can used to project batting average, slugging percentage, isolated power, K/AB, HR/AB, line drive rate, ground ball rate, etc.

Typical model optimization includes a total of 100,000 iterations of birth-death model sampling for each performance measure. The top five models are kept, as judged by the probability that the model truly represents the data (i.e., posterior model probabilities). The result is a system of five averaged and weighted models for each game variable.

The applications of such modeling are just beginning to be realized. Player Development staff can identify shortcomings in performance with precision, and better target more effective training needs. Operations Staff can use the information to safely decide when players should progress from level-to-level. Perhaps most importantly, Scouting Staff can use these data and projections to identify and procure talent earlier.

For more information or to schedule a consultation for your organization, please contact me via email (ben@motusglobal.com) or Twitter (@benhansen9).


Ben Hansen is the CTO at Motus Global. He specializes in movement analysis software development using wearable technology and motion capture cameras. Prior to Motus, Ben was a Ph.D. candidate at Marquette University (Biomedical Engineering) and a biomechanics engineer for the Milwaukee Brewers, helping pitchers, batters, and front office staff reduce injury risk and enhance on-field performance of players. His goal is to help people promote health and wellness through holistic performance data.