A rule induction framework for the determination of representative learning design in skilled performance

Area of Science:

• Sports Science
• Biomechanics
• Performance Analysis

Background:

• Representative learning design is crucial for effective sports training, aiming to replicate real-world performance conditions.
• Measuring and analyzing representative learning design can enhance training environments and athlete performance evaluation.
• Australian Rules football (AF) training can benefit from models that identify representative kicking events.

Purpose of the Study:

• To develop working models for determining high-frequency, representative kicking events in Australian Rules football.
• To apply rule induction techniques to analyze kicking constraints and outcomes.
• To provide a quantifiable approach for assessing the representativeness of sports practice.

Main Methods:

• Utilized a dataset of 9005 kicks from the 2015 Australian Football League season.
• Categorized and analyzed kicks based on pressure, distance, possession, time, velocity, and target.
• Employed the Apriori algorithm for rule induction to build predictive models.

Main Results:

• Developed a model with 10 rules identifying common kicking constraints (support 0.15-0.22, confidence 0.63-0.84).
• A second model achieved 70% classification accuracy for effective and ineffective kicks.
• Identified key constraints influencing representative kicking events in AF.

Conclusions:

• The study provides a novel measurement approach for quantifying sports practice representativeness.
• Findings are directly applicable to refining training drills in Australian Rules football and other team sports.
• This framework aids in creating more ecologically valid training environments for athletes.