EVOLVING
THE SIGNAL.
Success in high-variance environments isn't a snapshot; it's a process. I spent the 2025 season iteratively refining my feature weights and ensemble parameters, eventually securing the #1 ranking on the industry-standard leaderboard.
The Optimization Loop
Week 12
Top 7
Week 13
Top 13
Week 14
Mid-Pack
Post-Season
#1 Overall
CONTINUOUS
IMPROVEMENT.
Machine learning isn't "set and forget." Throughout the season, I monitored the model's Mean Squared Error (MSE) and Bias. By identifying a late-season drift in Week 14, I recalibrated the ensemble for Bowl Season, resulting in a 66.7% accuracy rate that beat both the Peer Average and the Vegas Line.
Peak Accuracy
66.7%
Market Edge
+4.5%
Post-Season Optimization Peak
CBB Production Pipeline
Feature-Engineered Ensembles
My College Basketball model applies a Random Forest Regressor to a high-dimensional feature set. I prioritize "Edge Metrics"—the delta between home/away efficiency ratings—to predict spreads with institutional-grade precision.
CBB Accuracy
Daily Automated Inference
Architecture
800-Estimator Ensemble
Data Logic
- Adj. Efficiency (SRS)
- Effective Field Goal %
- Possession Delta
- Temporal Decay Weights
Engineering Standards
The SOAR engine is built for Production Integrity, ensuring that data ingestion, cleaning, and modeling are decoupled and scalable.
Utilizing Pandas for vectorized operations and Scikit-Learn for high-performance machine learning.
Rigorous temporal validation ensures predictions are made only on data available at the time of the game.
Core Values
- Iterative Optimization
- Statistical Rigor
- Market Benchmarking
Engineered to Outperform.
Project Verification
DATA. PROVEN.
View the live leaderboard and tracking history via the links below.
Final Rank | #1 Overall (Bowls)