Value Verdict: Computational Risk Modeling and Inference Platform

Value Verdict is a self-initiated full-stack application designed for real-time analysis of sports market and finding positive expected value opportunities. Built with a focus on robust backend processing and interactive frontend tools, this project demonstrates implementation of advanced algorithms for data analysis, risk simulation, and real-time visualizations. The core emphasis is on technical implementation: efficient data handling, model calibration, and dynamic charting-leveraging Python for backend computations, SQL for database management, and React/TypeScript for the user interface.

Key Technical Implementations

Implemented a Python pipeline to ingest and preprocess 220k+ events (odds, outcomes, history). The Shin method was integrated to infer implied probabilities and calibrated on the dataset; calibration charts were produced to validate results.

Staking and Risk Simulation Algorithms

Core backend implements Kelly criterion (with fractional variants) and large-scale Monte Carlo simulations to project bankroll growth, standard deviation, and ruin probabilities. Vectorized operations were used to scale simulations efficiently.

Interactive Dashboards and Visualizations

The frontend, built in React/TypeScript, features dynamic dashboards where users can query events and view projections. Integration between backend APIs (Flask-based) and frontend components uses hooks for state management and real-time updates. Key visualizations include:

• Bankroll vs. number of bets: chart showing the bankroll as a function of the amount of bets, highlighting the growth over time.
• Bell Curve of Bankroll: visualizes the current bankroll compared to the expectations.
• Odds Distribution Histogram: bins odds data to reveal market biases.
• Win-Loss Breakdown Pie Chart: outcome breakdown with percentage annotations.
• Stakes Distribution Histogram: visualizes spread of computed stake sizes.
• Stakes vs. Odds Scatter Plot and Stakes vs. Log(Odds) Scatter Plot.

Results

During testing, the system proposed over 1300 sport events, and the total ROI was 6.8672%, while performing in the 50.86 percentile.

Tech Stack

• Backend: Python (modeling, ETL)
• Database: SQL (PostgreSQL)
• Frontend: React / TypeScript
• Modeling: Monte Carlo Simulation
• Algorithms: Shin Method, Kelly Criterion
• Visualization: Matplotlib / Seaborn, interactive SVGs
• Other: Git, FOOTBALLAPI