Automated MLB Underdog Predictor
Budget: $30 – $250 USD
I’m looking for a Python-based, fully automated model that can predict when an MLB underdog (money-line between +130 and +250) is likely to win outright. Each morning the script should pull the latest numbers from a reliable third-party sports data API, retrain or refresh projections, and output the day’s recommended plays without any manual intervention.
Key details
• Data feed: I’m leaning toward reputable third-party sports data APIs. If you feel an official or open-source feed would strengthen the model, let me know, but please assume the main pipeline comes from a paid provider; I’ll cover reasonable subscription costs.
• Workflow: The entire process—data ingestion, feature engineering, model training, prediction generation, and results export—must run on a schedule (cron, cloud function, or similar).
• Accuracy goal: Sustain at least 50 % hit rate on qualifying underdogs when measured over a meaningful sample size. Please include a back-test to demonstrate historical performance.
• Output: Daily list (CSV, JSON, or Google Sheet) of money-line underdogs meeting the criteria, each with win probability, implied edge, and any key model notes.
• Stack: Python with common ML libraries (pandas, scikit-learn, XGBoost, etc.) and robust logging/error handling.
Deliverables
1. Clean, well-commented Python codebase with setup instructions
2. Automated scheduler or deployment script (Docker, AWS Lambda, or your proposed solution)
3. Documentation covering data sources, feature set, model methodology, and how to adjust odds bands or thresholds
4. Back-test report validating the 50 %+ accuracy target
5. One brief hand-off session to walk through the system and API billing setup
Acceptance criteria
• End-to-end job runs unattended for at least one week in a test environment
• Daily predictions only include games where the closing line at pull time is +130 to +250
• Historical back-test shows ≥ 50 % accuracy on those picks over multiple seasons
• Code passes a quick review for readability, modularity, and reproducibility
If this sounds like a challenge you’d enjoy, tell me how you’d tackle the feature set and which third-party API you recommend so we can get started right away.
Key details
• Data feed: I’m leaning toward reputable third-party sports data APIs. If you feel an official or open-source feed would strengthen the model, let me know, but please assume the main pipeline comes from a paid provider; I’ll cover reasonable subscription costs.
• Workflow: The entire process—data ingestion, feature engineering, model training, prediction generation, and results export—must run on a schedule (cron, cloud function, or similar).
• Accuracy goal: Sustain at least 50 % hit rate on qualifying underdogs when measured over a meaningful sample size. Please include a back-test to demonstrate historical performance.
• Output: Daily list (CSV, JSON, or Google Sheet) of money-line underdogs meeting the criteria, each with win probability, implied edge, and any key model notes.
• Stack: Python with common ML libraries (pandas, scikit-learn, XGBoost, etc.) and robust logging/error handling.
Deliverables
1. Clean, well-commented Python codebase with setup instructions
2. Automated scheduler or deployment script (Docker, AWS Lambda, or your proposed solution)
3. Documentation covering data sources, feature set, model methodology, and how to adjust odds bands or thresholds
4. Back-test report validating the 50 %+ accuracy target
5. One brief hand-off session to walk through the system and API billing setup
Acceptance criteria
• End-to-end job runs unattended for at least one week in a test environment
• Daily predictions only include games where the closing line at pull time is +130 to +250
• Historical back-test shows ≥ 50 % accuracy on those picks over multiple seasons
• Code passes a quick review for readability, modularity, and reproducibility
If this sounds like a challenge you’d enjoy, tell me how you’d tackle the feature set and which third-party API you recommend so we can get started right away.
Related categories:
PHP
JavaScript
Python
Software Architecture
Docker
Aws Lambda
Data Analysis
API Integration