Profiling the Physical Performance of Young Boxers with Unsupervised Machine Learning: A Cross-Sectional Study

Area of Science:

• Sports Science
• Biomechanics
• Exercise Physiology

Background:

• Mexico City hosts the largest boxer population in Mexico.
• The city pioneered a sports baccalaureate program with a boxing focus.
• Understanding young boxers' physical profiles is crucial for specialized training.

Purpose of the Study:

• To determine the physical-functional profile of young boxing athletes applying for a sports baccalaureate.
• To identify distinct physical characteristics within this cohort.
• To inform the development of targeted training strategies.

Main Methods:

• A cross-sectional study involving 227 young boxers (44F, 183M).
• Measurements included body mass, handgrip strength, countermovement jump height, punch velocity, and punch impact force.
• Unsupervised machine learning (K-Medoids clustering) was used for profiling.

Main Results:

• Two distinct, sex-independent groups emerged: Profile 1 (n=118) and Profile 2 (n=109).
• Profile 2 demonstrated statistically significant higher values in punch impact force and related ratios compared to Profile 1.
• Strength levels were the primary factor explaining data variation.

Conclusions:

• The identified physical-functional profiles can aid in differentiating training programs for young boxers.
• Assessment of isometric and applied strength in boxing-specific movements is recommended.
• This profiling can support sports specialization and athlete development.