Overtraining does not arise from a single bad day. That is why monitoring is difficult: heart-rate variability, sleep, motivation and performance can all change together, but not necessarily in the same direction.
A study published in 2026 followed 14 high-performance swimmers through three consecutive seven-day loading cycles.
Many signals, one complex picture
[bild id="XX"]
The researchers collected HRV from orthostatic tests, sleep data, subjective load, motivation, strength and swimming performance, among other variables.
They then used Explainable Artificial Intelligence – models designed not merely to produce black-box decisions, but to make the most influential variables visible.
This is where AI can add value
A coach can easily recognise individual trends. It becomes more difficult when several data sources fluctuate at the same time and the interactions between them change.
AI models could help identify patterns that are overlooked when data are viewed only in ordinary tables.
Fourteen athletes are not an oracle
The sample is small. A model that works within one training group will not automatically transfer to other age groups, performance levels or training systems.
That is exactly why explainability matters. Coaches should be able to understand why a system classifies a load state as unusual.
The future of monitoring may therefore not be an AI that tells the coach what to do. A more useful AI may say: these three signals have changed together in an unusual way – take a closer look.
Explainable AI matters especially for coaches
A warning without a reason creates little confidence. If a model instead shows that sleep, HRV and subjective load, for example, are all deviating from an athlete's individual pattern, the coach can ask targeted questions. Transparency itself therefore becomes a quality feature of monitoring.
SCIENTIFIC SOURCE
https://pubmed.ncbi.nlm.nih.gov/41348148/