FIFA World Cup Prediction Model

Monte Carlo simulation of FIFA tournament outcomes driven by Elo ratings with quantified uncertainty. 100,000 simulations of the full 48-team 2026 format: 12 groups → 32-team knockout → champion.

Quick Start

pip install numpy pandas
python sim.py

Output: data/wc2026_forecast_results.csv — 48 teams × 7 stage probabilities.

Model

μ = current Elo, σ = historical volatility.

1. elo_forecast.py reads 2010–2025 yearly Elo snapshots from eloratings.net, fits a detrended linear model per team, and outputs pred_elo (= last Elo, optimal point prediction) and sigma (= residual std, the uncertainty).
2. sim.py draws each team's Elo from N(pred_elo, sigma) every simulation, converts to expected goals via Poisson, and simulates group stage → knockout bracket with FIFA's official third-place assignment rules.
3. Stage probabilities are frequency counts over 100,000 independent Monte Carlo runs.

Full algorithm: MODEL.md. Adapting to other tournaments: ADAPTATION.md.

Files

File	Purpose
sim.py	Monte Carlo simulator (100k iterations)
elo_forecast.py	Per-team σ estimation from 2010–2025 history
data/elo_forecasts.csv	Pre-computed (μ, σ) for 48 teams
data/eloratings/	Yearly Elo snapshots (2010–2025 + 2026_graph)
data/wc2026_forecast_results.csv	Simulation output

Data Sources

• Elo ratings: eloratings.net — public TSV files, no API key required
• Tournament structure: FIFA World Cup 2026 regulations (12 groups, 32-team knockout)

License

MIT