Author: Dr. Jirka Poropudas
The real-time audio feedback delivered by the Spalding TF DNA app is anchored in an empirical baseline derived from over 30 million shots logged across the platform.
Based on the collected shot data, optimal target ranges have been solved for:
- Release time
- Shot arc (Entry angle)
- Spin rate
In the Spalding TF DNA app, the audio feedback is based on optimal shot parameters estimated from 30 million shots collected from app users.
How Real-Time Audio Feedback Works
When a shot is detected, the app processes the sensor data collected during the shot and the user is given closed-loop audio feedback if:
- The shot metric falls inside or outside the parameter's optimal target window.
- The user has enabled real-time audio feedback for that specific metric.
To analyze how audio feedback affects the improvement of the players who use the app, we conducted the following feedback study.
Study Data and Player Selection
To evaluate the isolated impact of audio feedback on motor skill acquisition, data was extracted from app usage between January 1, 2026, and May 25, 2026.
Strict inclusion criteria were applied to minimize confounding variables and ensure a homogeneous sample:
- Novel users only: Limited strictly to users who first used the app after January 1, 2026, eliminating historical training bias.
- Demographic uniformity: Limited to players under 21 years of age.
- Spatial control: Restricted to midrange shots executed from between 10 to 15 feet to eliminate the effect of variable shooting profiles.
Statistical Model
To account for varying usage patterns where individual users may take shots both with and without audio feedback active, we fitted a Generalized Linear Mixed-Effects Model (GLMM) to predict session-by-session shooting performance.
The model incorporates two key structural elements:
- Random intercepts and slopes: Accounts for baseline differences in individual skill levels at entry, as well as unique individual learning rates over time.
- Logistic link function: Models the natural learning curve of human motor skills, reflecting diminishing marginal returns (i.e., as a shooter's performance improves, additional percentage point gains require exponentially greater effort).
Key Variables Evaluated
The key independent variables evaluated in the model were:
- The cumulative number of shots taken without real-time audio feedback for the given metric.
- The cumulative number of shots taken with real-time audio feedback for the given metric.
Standardized 10,000-Shot Comparison
To present comparative outcomes cleanly in our case studies, model predictions were standardized to evaluate performance gains across a benchmark volume of 10,000 cumulative shot repetitions:
- Control Group (Group A): Model estimation for 10,000 cumulative shots executed without audio feedback (0% feedback exposure).
- Intervention Group (Group B): Model estimation for 10,000 cumulative shots executed with real-time audio feedback (100% feedback exposure).
Explore the Study Results
The results of the study are presented in more detail in the following documents: