A desktop app that treats soccer betting like a data science problem

What it does

ProphitBet is a Python desktop application with a Tkinter GUI that downloads historical soccer data from football-data.co.uk, scrapes upcoming fixtures from Footystats, and trains ML models to predict match outcomes. It supports neural networks, random forests, and ensemble methods, then exports predictions to CSV or Excel. The whole pipeline is wrapped in a point-and-click interface with light and dark themes.

The interesting bit

The app doesn’t just throw algorithms at the problem and hope. It includes a surprisingly thorough feature-engineering workshop: Boruta algorithm for feature selection, correlation analysis, variance filtering, and even logistic regression coefficients for interpretability. There’s a “Profit Balance” metric that claims to flag whether your selected matches are mathematically profitable — a rare nod to the actual economics of betting that most prediction toys ignore.

Key highlights

• Built-in data analysis suite: distributions, variance, correlation, descriptive stats, and Boruta feature ranking
• Cross-validation and holdout validation automatically applied during training to reduce overfitting
• Excel-like match table with search, filtering, and direct Excel export
• Scrapes upcoming fixtures from Footystats for forward-looking predictions
• “Explainable Models” as an explicit design goal, with tools for coefficient and impurity analysis

Caveats

• The README is heavy on installation instructions and light on model performance metrics or validation results
• Several recent releases focused on bug fixes and parser failures, suggesting the data pipeline can be brittle
• No discussion of odds integration or bankroll management beyond the single “Profit Balance” metric

Verdict

Worth a look if you’re a data-curious soccer fan who wants to experiment with feature engineering and model interpretability without writing code. Skip it if you’re seeking a proven profitable betting system — the app provides tools, not evidence that they work.