Skip to content

README.md

Latest commit

 

History

History
107 lines (77 loc) · 2.63 KB

README.md

File metadata and controls

107 lines (77 loc) · 2.63 KB

Analysis README

This folder contains the simulation script and Marimo notebooks for analyzing court assignment algorithms.

Setup (install dependencies)

Dependencies are declared in analysis/pyproject.toml.

cd analysis
uv venv
./.venv/bin/uv pip install .

Run the simulation (generate CSVs)

The simulation compares all four court assignment algorithms (Random Baseline, Monte Carlo, Simulated Annealing, and Conflict Graph) and generates comprehensive metrics.

Configuration

Create or edit analysis/data/config.json:

{
  "runs": 100,
  "rounds": 10,
  "playerCounts": [20],
  "numCourts": 4
}
  • runs: Number of simulation sessions per algorithm
  • rounds: Number of rounds per session
  • playerCounts: Array of player counts to test (e.g., [17, 18, 19, 20])
  • numCourts: Number of available courts

Run the simulation

cd analysis
npx tsx ./simulation/index.ts

Output

The simulation generates data in analysis/data/:

  • random_baseline/, mc_algo/, sa_algo/, cg_algo/: Per-engine results
    • summary.csv: Repeat pair statistics per simulation
    • pair_events.csv: Individual repeat pair occurrences
    • match_events.csv: Match outcomes with team strengths
    • bench_stats.csv: Bench fairness metrics
    • match_pair_summary.csv: Player pair interaction counts
    • player_stats.csv: Win/loss records per player
    • config.json: Configuration and aggregate statistics
  • comparison_summary.csv: Side-by-side engine comparison

Metrics tracked

  • Repeat pairs: How often the same team plays together across rounds
  • Balance: Team strength matching (based on simulated skill levels)
  • Bench fairness: Distribution of bench time and gaps between benches
  • Win distribution: Player win rates relative to skill levels

Run the notebook (editable)

cd analysis
./.venv/bin/marimo edit ./engine_analysis.py

Run the notebook (read-only app)

cd analysis
./.venv/bin/marimo run ./engine_analysis.py --host 127.0.0.1 --port 2786

Run all notebooks (with hot reload)

Serves all notebooks with automatic refresh on file changes:

cd analysis
uv run serve.py

Then open:

Export notebooks to static HTML

Export all notebooks without code cells (output only):

cd analysis
uv run export-html

Then prerender for the main app (from project root):

npx tsx script/prerender-notebooks.ts

This generates static HTML files in public/analysis/ that render without JavaScript execution.