KISS-BIS — AI for Elite Sports
2019–2020 Machine Learning Sports Science Small Data
KISS-BIS — Artificial Intelligence for Elite Sports in the Tension between Big and Small Data
The value of AI and machine learning in sports is widely acknowledged, yet practical adoption is slow. KISS-BIS investigated why — identifying methodological, structural, and communicative mismatches between AI research and elite sports practice.
Problem
Elite sports typically operates in a small-data regime: athletes are few, measurements are costly, and inter-individual variation is high. This contrasts sharply with the big-data assumptions underlying many modern ML methods. The project developed a research agenda and use-case demonstrators for bridging this gap.
Approach
- Systematic review of AI/ML applications in national and international elite sports, and from analogous domains
- Three use cases: training load modelling, competition diagnostics/game analysis, and performance diagnostics
- Transfer recommendations for making ML methods practically usable in sports organisations
My role
Contributed to the systematic review methodology and the data analysis pipeline for training load modelling use cases.
Partners
Prof. Alexander Asteroth (lead, H-BRS / TREE), University of Mainz, TU Munich. Funded by BISp (Bundesinstitut für Sportwissenschaft) and DOSB.