Training control through artificial intelligence
A recent study in the open access journal Nutrients shows how artificial intelligence (AI) is increasingly changing training management in endurance sports. The authors Gerasimos V. Grivas and Kousar Safari describe how AI uses data from wearables, metabolic measurements and training protocols to customise training, recovery and nutrition. The key message: away from rough training plans and towards precise, personalised recommendations - also relevant for sports such as mountain biking, where exertion and recovery fluctuate greatly.
AI systems now analyse heart rate variability, sleep quality, nutritional data and even glucose levels. This results in specific recommendations for action, such as when an intensive cycling day makes sense or when it would be better to do a relaxed rolling session. According to the authors, this approach is more accurate than traditional coaching methods or rigid training plans.
AI as a prediction tool
Traditionally, endurance training is mainly based on heart rate, lactate tests, subjective perception of exertion and experience. However, these methods usually only provide snapshots. AI, on the other hand, can analyse a lot of data simultaneously and derive trends and predictions from it - for example, how well you will feel the next day or how much energy you will need during a long training session or a race on difficult terrain.
The study shows that AI models are particularly good at:
- the forecast of recovery on the following day
- the estimation of metabolic thresholds without a laboratory test
- the optimisation of carbohydrate and fluid intake
- the creation of route-specific speed strategies, for example for long climbs
AI therefore not only describes the current situation, but also looks ahead.
Will AI replace traditional performance tests?
Deep learning models can now estimate lactate and ventilation thresholds quite accurately from non-invasive data such as heart rate and power output. For cyclists, this could mean determining training zones without having to go to the lab regularly. However, it remains to be seen whether AI will completely replace performance tests or merely supplement them.
Targeted control of nutrition in competition
A particularly practical area of application is competition nutrition. Continuous glucose meters, originally developed for diabetics, provide data that AI analyses in real time. This allows the carbohydrate intake during a race or a long tour to be adjusted in a targeted manner - for example when the intensity changes between climbs, flat sections and descents.
The models take into account individual reactions, environmental conditions and stress. The measurements are not yet perfect, which is why recommendations should always be subject to certain uncertainties.
Better recovery forecasts
A twelve-week study involving 43 endurance athletes demonstrated the performance of machine learning in predicting recovery. The algorithms analysed daily heart rate variability together with training load, sleep, nutrition and wellness parameters. They predicted morning recovery status and daily heart rate changes more accurately than simple reference models.
AI systems now integrate additional biometric signals such as resting heart rate, breathing rate, sleep architecture and skin temperature from wearable devices. These complex data sets enable more differentiated assessments of recovery processes.
AI can predict, among other things:
- How refreshed you feel the next morning
- Heart rate changes
- whether fatigue builds up
- Early warning signs of overtraining
These models are often more differentiated than the well-known "recovery scores" of many sports watches.
And data protection?
At the same time, AI raises questions about data protection. Wearables collect very personal data such as sleep, location or health values. It is often unclear how this data is stored or used. In addition, many AI models are difficult to understand and can be distorted if they have only been trained with data from elite athletes - which limits their informative value for recreational cyclists.
Outlook
The authors emphasise that AI can only fulfil its potential in endurance sports if diverse and high-quality data is used. Close collaboration between technicians, coaches and athletes is crucial. If used correctly, AI could make training more precise, safer and more effective for many cyclists - instead of just a small elite group.
Link to the Study
Kristian Bauer
Editor
