Real-Time Performance Prediction in Long-Distance Trail Running: A Practical Model Based on Terrain Difficulty and...

Area of Science:

• Sports Science
• Endurance Sports Analytics
• Performance Prediction Modeling

Background:

• Traditional performance prediction models for trail running often rely on lab tests or pre-race data, limiting practical, real-time application.
• Accurate prediction of marathon and ultra-trail race performance is crucial for strategic planning and injury prevention.

Purpose of the Study:

• To develop and validate a real-time predictive model for marathon and ultra-trail races using in-race data.
• To assess the model's predictive power across different sexes and race types.
• To provide a practical tool for athletes and coaches to monitor and optimize race strategy.

Main Methods:

• Analysis of 947 runners from the 'Trail Valle de Tena' event.
• Development of predictive equations using data from the first third of the race.
• Inclusion of variables such as weighted time (WTn), pacing variability (WTVn,n+2), and checkpoint percentile rank (CPRn).

Main Results:

• The developed model demonstrated strong predictive power, with an adjusted R-squared value greater than 0.95.
• The model effectively estimated total race time using only early-race data.
• Key variables (WTn, WTVn,n+2, CPRn) reflected a runner's ability to manage elevation and pacing consistency.

Conclusions:

• The real-time predictive model is highly applicable in field conditions, offering a practical alternative to lab-based methods.
• The model provides valuable insights for fatigue management and performance optimization during endurance races.
• Further validation in similar events is recommended to confirm its utility for training and competition planning.