PLForecast - Premier League Prediction Dashboard (v1)

A Monte Carlo simulation engine that predicts Premier League outcomes with interactive visualizations. Currently tracking the 2025-26 season with 20,000 simulations per run.

Features

• Title Probability - Real-time chances for each team to win the league
• Top 4 (UCL) Probability - Champions League qualification odds
• Relegation Battle - Drop zone probabilities with full distribution
• Expected Points Table - Projected final standings with confidence intervals
• Position Distribution Heatmap - Probability of finishing in each league position
• Team Deep Dive - Detailed breakdown for any selected team
• Smart Caching - 24-hour cache with instant reload and force refresh

Current Predictions (as of Feb 2026)

Team	Title %	Top 4 %	Relegation %	Exp Points
Arsenal	83.7%	99.9%	0.0%	77.5
Man City	9.2%	95.6%	0.0%	70.0
Aston Villa	6.6%	92.4%	0.0%	69.3
... full table in dashboard

Tech Stack

• Frontend: Streamlit + Plotly
• Backend: Python (pandas, numpy)
• Simulation: Monte Carlo (20k runs)
• Rating System: ELO-based with home advantage
• Caching: JSON-based with metadata validation
• Data Sources: football-data.co.uk, BBC Sport

Project Structure:

./
├── cache/   #auto generated cache
│   ├── elo_ratings_2526.csv
│   ├── latest_simulations.json
├── data/
│   ├── bbc_raw.html
│   ├── EO_2526_raw.json   #json file
│   ├── E0_2526.csv
│   ├── elo_ratings_2526.csv    #elo ratings
│   ├── power_rankings_2526.csv   #elo ratings sorted
│   ├── remaining_fixtures2526.csv   #remaining fixtures in 25-26 season
│   ├── simulation_results_2526.csv   #results from monte-carlo simulation
│   └── summary_table2526.csv
├── viz/
│   ├── match-level-eda.ipynb
│   ├── table-level-eda.ipynb
│   ├── mle.py   #.py file copy of match-level-eda.ipynb to import draw-rate variable
│   ├── power-rankings-viz.ipynb
│   ├── betting-odds-analysis.ipynb   #currently empty, will revisit in v2
│   └── README.md   #readme for more details about data EDA and visualization
├── src/
│   ├── config.py
│   ├── elo.py #generate elo's
│   ├── elo_run.py #generates elo's
│   ├── fetch_data.py   #gets data from website
│   ├── cache_utils.py   #caches data
│   ├── pipeline.py.   #define simulation pipeline (parameters, etc.)
│   ├── simulation.py   #contains details about running the monte-carlo simulations (main logic here)
│   ├── table.py    #defines function for updating our simulated league table
│   ├── remaining_fixtures.py       #loads remaining fixtures
│   ├── load_data.py    #loading season data from `EO_2526.csv`, power rankings, and current league table
│   └── README.md   #readme for more details about `src`
├── app.py   #Main Streamlit Dashboard
├── run.py    #Entry point for simulations
├── cache_manager.py   #saving, loading, & managing cacheing of results
├── data_processor.py   #processes simulation results data for streamlit + plotly visualization
├── LICENSE
├── requirements.txt
└── README.md

Descriptions:

• All analysis notebooks are located in notebook-viz/.
• Logic and data-fetching will be located in src/.

How It Works

1. ELO Rating System

• Each team starts with a base ELO rating of 1500
• Ratings update after every match based on result and margin
• Home advantage factor built into calculations

2. Monte Carlo Simulation

• Takes current ELO ratings and remaining fixtures
• Simulates each match 20,000 times using probability distributions
• Updates league table after each simulated match
• Records final positions across all simulations

3. Probability Calculations

• Title %: Times team finishes 1st / total simulations
• Top 4 %: Times team finishes 1st-4th / total simulations
• Relegation %: Times team finishes 18th-20th / total simulations
• Expected Points: Average points across all simulations

4. Caching Strategy

• Results cached for 24 hours (configurable)
• Cache invalidated when new match data arrives
• Manual override with "Force Refresh" button

Local Installation

# Clone repository
git clone https://github.com/yourusername/PLForecast.git
cd PLForecast

# Install dependencies
pip install -r requirements.txt

# Run dashboard
streamlit run app.py

# or

python refresh_and_run.py --launch-app

Future Enhancements in v2:

• Live data fetching via API (football-data.org)
• Automated nightly simulation runs
• Poisson model as alternative to ELO
• Betting odds integration with value detection
• Historical accuracy tracking
• Head-to-head match predictor
• Export functionality (PDF/CSV)
• Mobile-responsive design

---

NOTICE:

This project uses football data for educational and demonstration purposes.

Important Disclaimers:

• This project is not affiliated with, endorsed by, or connected to the Premier League
• Club names, logos, and trademarks are property of their respective owners
• Predictions are based on statistical models and should not be used for betting