How BetEdge AI Makes Predictions
A transparent look at our prediction engine, from data ingestion to cryptographic verification. No black boxes.
Two-Layer Prediction System
ML Primary Path
Trained XGBoost and LightGBM models loaded via a ModelRegistry from serialized .pkl files. Each sport has its own model trained on historical game data, line movements, and team performance metrics.
Model versions are tracked with model_version field (e.g., ml-trained-nba-v2).
Rule-Based Fallback
When ML models aren't available or confident enough, a rule-based system using team tiers and historical matchup data provides predictions. These always include a disclaimer in the reasoning.
Model versions: statistical-v1. Easily identifiable in the track record.
Feature Engineering
Our models ingest 24+ features per game across six categories. Each feature is engineered to capture predictive signal while minimizing overfitting.
Team Performance
- Win/loss record (last 10, 20, season)
- Points scored & allowed averages
- Home/away splits
- Strength of schedule
Market Data
- Opening & closing lines
- Line movement direction & velocity
- Sharp money indicators
- Consensus pick percentages
Situational Factors
- Rest days between games
- Travel distance
- Back-to-back detection
- Injury report integration
Historical Patterns
- Head-to-head matchup history
- ATS (against the spread) trends
- Over/under tendencies
- Division/conference performance
Advanced Metrics
- Offensive/defensive efficiency ratings
- Pace & tempo adjustments
- Expected points models
- Four-factor analytics (basketball)
Model Ensemble
- XGBoost gradient boosting
- LightGBM tree ensemble
- Model confidence calibration
- Cross-validation stability scores
Confidence Calibration
Raw model probabilities are calibrated so that a “70% confidence” prediction wins approximately 70% of the time. We use isotonic regression on holdout data to align predicted probabilities with observed outcomes.
High Confidence (80%+)
Strong model agreement with clear market inefficiency. Fewer picks, higher expected edge.
Medium Confidence (60-79%)
Solid directional signal with some uncertainty. Most daily picks fall in this range.
Lower Confidence (50-59%)
Slight edge detected. Only surfaced when combined with positive expected value (+EV).
Cryptographic Verification
Every prediction is SHA-256 hashed before the game starts to guarantee we can't alter picks retroactively. This is how we prove our track record is real.
1. Hash Generation
When a prediction is created, we compute SHA256(predictedOutcome + confidence + predictionType + gameId + modelVersion) and store the hash permanently.
2. Immutable Storage
The hash is stored alongside the prediction. Once the game starts, neither the prediction nor the hash can be modified.
3. Public Verification
Anyone can verify any prediction via our public API: GET /api/public/verify/[predictionId]. The response includes the stored hash and a recomputed hash for comparison.
See It in Action
Browse our verified public track record or start getting predictions today.