In the lab
Modeling athlete and team performance
Building a dataset and prediction pipeline with a cross-country coach to explore individual performance and head-to-head team outcomes.
- Status
- Early research
- Signal
- Dataset design · prediction pipelines · early ML testing
- Focus
- Machine learning · Data pipelines · Evaluation · Sports
The question
What can historical race data say about an individual athlete’s next performance—and how should those estimates combine when two teams meet?
I am working with a cross-country coach to build the dataset and prediction pipelines needed to investigate both questions. The project is early, and this page records the reasoning rather than presenting preliminary results as a finished product.
Current work
- Define a dataset that preserves athlete, course, meet, and temporal context.
- Establish useful baselines before increasing model complexity.
- Evaluate individual predictions separately from head-to-head team outcomes.
- Turn coaching knowledge into testable features without encoding assumptions that the data cannot support.
Early signal
Initial machine-learning tests are promising enough to continue the work, but the evidence is not mature enough for a performance claim.
Why show unfinished work
Architecture is visible before a model is impressive. Dataset boundaries, leakage prevention, evaluation design, and the translation between domain expertise and measurable signals are the durable parts of the project.