This project explores how combinatorial optimization and game theory can enhance hero selection strategies in the team-based shooter Marvel Rivals. By modeling hero matchups using real-world performance data, we analyze how optimal team compositions can be constructed under various competitive conditions.
We build a game-theoretic model around team selection in Marvel Rivals, treating each 6-hero lineup as a discrete strategy. A payoff matrix is derived from publicly available win rate data, and several optimization problems are formulated to reflect realistic scenarios such as:
- Optimal counter-teaming
- Banning mechanics
- Ultimate usage maximization
- Role-balanced team compositions
- Team-Up synergy optimization
| File | Description |
|---|---|
dataImport_Script.py |
Script to scrape and update the latest win rate and matchup data from rivalsmeta.com |
main.ipynb |
Full implementation of all optimization problems discussed in the project |
MarvelRivals_WinRate_Matrix.csv |
Raw win rate matrix (unprocessed) |
MarvelRivals_NumMatches_Matrix.csv |
Matrix of total match counts between heroes |
MarvelRivals_Payoff_Matrix.csv |
Normalized payoff matrix used in optimization |
- Counter-strategy generation using linear and binary programming
- Minimax strategies based on von Neumann’s theorem
- Banning constraints and feasible team construction
- Multi-objective optimization combining ultimates and win rates
- Role-based filtering and synergy-aware (Team-Up) team formation
All matchup data was obtained from:
rivalsmeta.com — the most comprehensive source of performance analytics for Marvel Rivals.
- Update data: Run
dataImport_Script.pyto scrape the latest character matchup data. - Solve problems: Open
main.ipynbfor the full suite of optimization tools and demonstrations.
March 17, 2025