Predicting Basketball Shot Outcome From Visuomotor Control Data Using Explainable Machine Learning.

Area of Science:

• Sports Science
• Motor Control
• Human Performance

Background:

• Quiet eye (QE) is a visual fixation on a target before a critical action, linked to enhanced performance.
• While QE is trainable, its direct predictive power for performance outcomes remains under investigation.
• Understanding QE's predictive capacity is crucial for developing effective training interventions in sports.

Purpose of the Study:

• To predict basketball shot outcomes (make or miss) using visuomotor control variables.
• To investigate the predictive accuracy of different QE parameters and other shot-related variables.
• To determine if QE can serve as a direct performance predictor in basketball.

Main Methods:

• A decision tree classification approach was employed to predict shot outcomes.
• Twelve basketball athletes performed 200 shots from various on-court locations while wearing mobile eye-tracking glasses.
• Eight predictors, including shot location, arm extension time, and QE metrics (onset, offset, duration), were modeled using decision trees and random forests.

Main Results:

• Decision tree models accurately predicted over 66% of successful shots and over 50% of missed shots.
• Relative quiet eye duration emerged as the primary predictor of shot success.
• Successful shots were associated with relative QE durations exceeding 18.4%.

Conclusions:

• Quiet eye duration is a significant predictor of basketball shot success.
• Training interventions aimed at prolonging QE duration beyond 18% may enhance shooting performance.
• These findings support the use of QE as a measurable target for improving athletic performance.