The conceptual components of the project are as follows:

  1. Development of an app for recording of training load, individual stress-/health-/recovery level and providing (data-driven) feedback/output to athletes and coaches;
  2. Incorporation of data from external devices (wearables, biomarkers);
  3. Incorporation of external measurement devices (ergometer, boat);
  4. Regular standardized test sequences to measure (all-out) performance;
  5. Statistical modelling and machine learning based on data from a)-d) in order to predict the effect of training and potential performance on individual basis;
  6. Translation of statistical predictions into informative and useable indicators for athletes and coaches.

The project will be performed on a modular basis with the following working groups:

Working Group "Data & App"

  • Development of an app for manual recording of training data and individual wellbeing/health data;
  • Integration of biomarkers;
  • Integration of APIs (Application Programming Interfaces) of external devices (e.g. wearables);
  • Development of explorative statistics and visualizations of indicators (based on results from working group "Data Science") for feedback to coaches and athletes;
  • Setting up IT infrastructure for recording, storage and administration of data, ensuring regulations for data privacy and protection
  • Supervision of athletes' compliance 

Working Group "Data Science"

  • Development of appropriate statistical approaches  of machine learning (e.g. pattern recognition, classification methods, artificial neural networks, filtering, dimension reduction, multivariate time series techniques) in order to
    • quantify internal training load
    • identify relevant variables/markers serving as reliable proxies for internal load (variable selection)
    • predict the effects of certain training stimulus conditionally on athlete-specific states and training load records
    • quantify the relationships between training load, athlete-specific factors and (potential) all-out performance
    • predict potential all-out performance based on (sequences of) sub-tests
  • Validation and potential combination of competing statistical approaches
  • Combination of machine learning with structural approaches from sport sciences 
  • Producing suitable indicators and recommended actions as feedback to athletes and coaches (to be visualized through the app)

Working Group "Sports Science"

  • Adaption and implementation of the "Performance Potential Double Model (PerPot DoMo)" for the current use case 
  • Using "PerPot DoMo" as a benchmark model for alternative approaches to be developed in the Working Group "Data Science"
  • Development of appropriate hard-/software for precise measurement of power in the boat and on the ergometer; adaption of ergometers for regular all-out test sequences
  • Development of approaches for the quantification of training load caused by strength training

Working Group "Sports Medicine"

  • Identification and implementation of appropriate bio markers for the quantification of internal load
  • Identification of appropriate wearables and corresponding APIs 
  • Medical supervision 
  • Integration of aspects from research on sleep quality and nutrition
  • Medical input for other working groups