EPL-Analyst

• 

  EPL-Analyst

Inspiration

I was interested in understanding the efficiency of bookmakers' odds in football betting, and also to find out if there are hidden opportunities that could be uncovered

What it does

This project analyzes bookmakers' odds and compares them with probabilities from the machine learning model. This will help identify fair, overpriced, and value bets

How we built it

To achieve this, I gathered over 2000 matches from 6 seasons in the English Premier League (from 2018 to 2024). I cleaned the data and structured it properly so that it can be used in this project. Then I converted odds into implied probabilities and engineered features such as Overround and match favorites.

After that, I trained a multinomial logistic regression model using a chronological train-test split to simulate real-world prediction sessions.

I was able to achieve all these with the help of the Zerve AI tool

Challenges we ran into

One of the key challenges I faced was handling issues like inconsistent team names and ensuring stable model training. Some teams did not have consistent naming across bookmakers

Accomplishments that we're proud of

1. I built a complete end-to-end machine learning system from raw EPL data to a fully deployed interactive web interface

2. Successfully trained a multinomial logistic regression model and achieved approximately 58% accuracy.

3. I designed a value bet detection system that identifies potential market inefficiencies

4 I implemented a calibration layer to analyze how well predicted probabilities match real-world outcomes

1. I deployed the project as a working frontend interface

What we learned

Through this project, I learned how to work with real-world data, engineer meaningful features from probabilities, evaluate model performance, and deploy a machine learning model as an interactive frontend.

What's next for EPL-Analyst

1. Experiment with more advanced models (e.g., Gradient Boosting, XGBoost) to capture nonlinear patterns in match outcomes

2. Expand beyond the EPL to include other major leagues and international competitions

Built With

• css3
• fastapi
• html
• javascript
• pandas
• python
• scikit-learn
• uvicorn