Identifying Key Training Load and Intensity Indicators in Ice Hockey Using Unsupervised Machine Learning.

Area of Science:

• Sports Science
• Biomechanics
• Exercise Physiology

Background:

• Accurate monitoring of training load (TL) and intensity is crucial for optimizing performance and preventing injuries in elite athletes.
• Ice hockey involves complex movements and high-intensity bursts, necessitating reliable methods to quantify physical demands.
• Previous research has explored various metrics, but a comprehensive understanding of their interrelationships in ice hockey is lacking.

Purpose of the Study:

• To identify key training load (TL) and intensity indicators in ice hockey.
• To examine the correlations between accelerometer-derived and heart rate-derived TL and intensity metrics.
• To explore the potential redundancy of using multiple monitoring variables.

Main Methods:

• Collected practice and game data from 17 elite Danish ice hockey players over a four-week competitive period.
• Utilized wearable 200-Hz accelerometers and heart rate monitors to record data.
• Applied within- and between-individual correlation analyses and K-means++ cluster analysis.

Main Results:

• Large to almost perfect correlations (r = 0.69–0.99) were found between various accelerometer- and HR-derived TL indicators.
• No significant correlations were observed between accelerometer- and HR-derived intensity indicators.
• Cluster analysis identified distinct groups based on between- and within-loadings, highlighting specific relevant variables for different analyses.

Conclusions:

• Multiple TL and intensity variables in ice hockey exhibit significant redundancy.
• Specific metrics like session duration, accelerations, decelerations, and heart rate parameters are key indicators.
• Understanding variable redundancy can inform more efficient monitoring strategies for ice hockey players.