Horse racing handicapping automation python-MVP

Project Brief: Horse Racing Handicapping Automation – MVP Project Title: Build an MVP Python script to automate horse racing handicapping using Thoro-Graph, Brisnet, and EquinEdge data. Overview / Goal I have developed a working handicapping model in Python (handicap_race_v39) that ranks horses for a race using normalized metrics from EquinEdge (Win%, GSR), Beyer figures, projected Thoro-Graph figures, and bonus signals (Thoro-Graph patterns, trip quality, hidden/sneaky good performances). The goal of this project is to create a Minimum Viable Product (MVP) that automates the end-to-end process: Input: A full race card’s documents (Thoro-Graph PDF + Brisnet PDF + EquinEdge screenshots) Output: Ranked horse selections per race with transparent bonus explanations. This is the cheapest realistic path — I want a functional working script first, not a polished GUI or production-grade system. Current State (What Already Exists) • A verified core handicapping function (handicap_race_v39) that normalizes inputs and produces ranked selections with estimated win probabilities. • Proximity-based bonus logic that assigns Thoro-Graph patterns and trip comments to the correct horse. • Bonus configuration values (e.g., Top-Pair-Top = +18, troubled but strong trip = +6 to +10, hidden trip = +12, X/bounce = -8). • Human-readable bonus summary logic. I will provide all existing code to the developer. MVP Scope (Cheapest Realistic Path) Build a single, reliable Python script that can: 1. Accept a race card’s documents (one Thoro-Graph multi-page PDF, one Brisnet PDF, and multiple EquinEdge screenshot images). 2. Extract key structured data: • Horse names • Today’s projected Thoro-Graph figure (or relevant pattern) • Beyer figures • EquinEdge Win% and GSR • Thoro-Graph patterns (Top-Pair-Top, Pair-Pair-Pair, X, bounce, etc.) • Trip quality / hidden trip signals 3. Run the existing handicap_race_v39 model with proximity-based bonuses. 4. Output clean ranked selections per race, including: • Rank • Horse name • Composite score • Estimated win probability • Key bonus explanations (why the horse received positive or negative bonuses) Key Deliverables • One working Python script (or small set of scripts) that processes a full race card. • Clear output in CSV and/or readable text format. • Basic documentation on how to run the script. • The script should handle the most common cases cleanly (even if it needs occasional manual help on very difficult pages). Technical Preferences • Python 3 (pandas, numpy, etc.) • Use of AWS Textract is acceptable for PDF parsing (I can provide AWS access or the developer can suggest alternatives). • The existing code I provide should be used as the foundation for the scoring engine. • Keep it simple and maintainable — this is an MVP. Out of Scope (to keep cost down) • Web interface or GUI • Fully automated daily processing / scheduling • Perfect accuracy on every single page (some manual review or overrides are acceptable in MVP) • Back-testing framework • Advanced machine learning models Success Criteria • The script can process a complete race card (Thoro-Graph + Brisnet + EquinEdge) and produce ranked selections for all races. • Bonus logic is applied at the horse level (not race level). • Output is clear enough that I can understand why each horse received its ranking and bonuses. • The script runs reliably on new race cards with reasonable accuracy. Timeline & Budget Guidance (Cheapest Realistic Path) • I am looking for the most cost-effective realistic solution, not the most polished version. • Realistic budget range for this MVP: $3,500 (depending on experience and location). • Timeline: 3–5 weeks is acceptable. How to Apply Please include the following in your proposal: 1. Your relevant experience with PDF parsing, OCR, or document automation (especially dense tables or racing/sports data). 2. A short description of how you would approach the Thoro-Graph PDF parsing challenge. 3. Your proposed timeline and total cost for the MVP as described. 4. Any questions you have about the existing code or scope.