Coverage
3 sportsNFL, NBA, and MLB pipeline adapters.Calibrated sports totals research
ProEdge Analytics
A multi-sport modeling pipeline for comparing model distributions with market totals—without presenting a black-box pick feed.
- Role
- Machine learning · API · Observability
- Year
- 2025
- Repository
- Public source ↗
01 / Interactive system slice
Inspect the workflow
Public sandbox · real engine with clearly labeled cached and preview statesPEPROEDGE
LIVE TOOL PUBLIC SANDBOX
MODEL v2.4 · CALIBRATEDDETERMINISTIC PREVIEW
Starting the free-tier service…
| Matchup | Game date | Sport | Line | Model | Conf | Out | Edge |
|---|---|---|---|---|---|---|---|
| Phoenix @ Denver | NBA | 226.5 | OVER 64% | 28% | 0 | OVER 226.5 +7.2% | |
| Buffalo @ Miami | NFL | 48.5 | UNDER 61% | 22% | 1 | UNDER 48.5 +5.4% | |
| Seattle @ Houston | MLB | 7.5 | OVER 57% | 14% | 0 | OVER 7.5 +3.1% |
Deterministic preview · selected matchup
Phoenix @ Denver
Market date ·Model signalOVER 226.5+7.2% model edge
- Market line
- 226.5
- Over
- 64%
- Under
- 36%
- Confidence
- 28%
- P(over) CI
- 54–73%
- Players out
- 0 Away 0 · Home 0
- Line movement
- 0.0
- Team inputs
- Baseline No live team context returned
LINE MATRIX
Threshold detail is unavailable for this cached row.
Data sourceDeterministic previewBoard scanSchedule precisionProvider market date; exact tipoff not supplied
Model estimates are research outputs, not a profitability claim.
02 / Engineering evidence
Claims with boundaries
Derived from the local implementation and offline artifacts.Feature surface
200+Pace, efficiency, context, and market-derived inputs.Validation
Walk-forwardTime-ordered evaluation designed to avoid future leakage.Problem
Why this exists
Point forecasts hide uncertainty, while random train/test splits overstate performance on time-dependent sports data.System
How it is bounded
The pipeline builds time-safe features, ensembles tree models, calibrates probabilities, and monitors drift. The board foregrounds the interval and the market comparison rather than a promise of profit.Outcome
What the demo proves
A research surface that makes uncertainty, model-market disagreement, and evaluation methodology inspectable.Implementation surface
FastAPIPostgreSQLXGBoostLightGBMGrafana
Building reliable data products?