Opponent-aware lineup optimizer
Constrained decision-support prototype
Turn roster constraints, uncertainty, and an opponent’s strengths into an explicit optimization problem.
- Status
- Local research prototype
- Kind
- Data & geometry
- Stack
- Python · Monte Carlo · Bipartite matching · Backtesting
BUILD NOTES / Local research prototypeOpponent-aware lineup optimizer
Project overview · not an interface capture
Constrained decision-support prototype
Python / Monte Carlo / Bipartite matchingProject overview · not an interface capture
The problem
The strongest individual projections do not necessarily produce a legal lineup or the best category matchup. Eligibility and uncertainty need to be part of the decision.
What was built
- Enumeration of eligible player subsets with bipartite matching to validate position assignments.
- Monte Carlo estimates of win, tie, and loss probabilities across nine categories.
- A local decision dashboard and limited historical replay for examining recommendations.
What I learned
A recommendation should show both feasibility and uncertainty. Historical replay helps find mistakes but does not establish future performance.