Roadmap
What we've built, what we're building, and where we're headed. BakerPicks is actively developed.
Three-Stage Model Pipeline
Game model, minutes model, and 7 per-stat rate models feeding into Monte Carlo simulation.
10,000 Monte Carlo Simulations
Joint sampling of minutes and per-36 rates with stat-specific noise distributions for full probability outputs.
Automated Daily Calibration
Nightly postgame pipeline: grade predictions, Platt recalibration, adaptive bias correction, drift detection.
SHAP Explainability
"Why This Bet" cards showing the top 3 feature drivers behind every prediction.
Model Version Registry
Full version control for all 9 models with staged promotion, quality gates, and one-command rollback.
Closing Line Value Tracking
Automated CLV computation: capture closing lines 15 min before tip, compare to bet placement odds.
Market Signal Analysis
Cross-book line comparison: consensus lines, outlier detection, sharp money signals, stale line detection.
BakerPicks.com Public Website
Public transparency dashboard, daily auto-published metrics, full model accountability.
Daily Auto-Published Transparency Reports
Automated pipeline pushing calibration, CLV, and performance data to BakerPicks.com every night.
Historical Backtest Explorer
Interactive tool to explore model performance across historical date ranges, stat types, and confidence tiers.
Email Performance Digest
Weekly email summarizing model performance: win rate, CLV trends, notable model changes, top segments.
Same-Game Parlay Optimizer
Public-facing correlation-aware SGP builder using the joint Monte Carlo samples.
Live In-Game Projections
Real-time projection updates based on current box score pace and game flow.