Behind the Predictions
Football Methodology
Data Sources: We use API-Football.com as our primary data source (1,180+ leagues worldwide), with Understat as a fallback for xG data in major European leagues. All data is validated and processed through our statistical pipeline.
Model: Poisson-based probability modeling with Dixon–Coles correlation for low-scoring matches, opponent‑adjusted strength ratings, and a recency‑weighted blend (exponential decay).
Coverage: Teams from any league with sufficient recent match data (typically 15–20 matches) can be analyzed. The system automatically spans multiple seasons to gather historical data.
Data Quality: We prioritize Expected Goals (xG) when available, falling back to observed goals for leagues without xG coverage. Recent matches are weighted more heavily than older ones.
Volatility Index
The Volatility Index is a stress-test for the model's confidence. It specifically tests the **structural integrity of the Handicap Line**—not just the raw win probability.
Why Test the Handicap?
We identify the Asian Handicap line closest to a 50/50 probability (the 'Main Line') and simulate adverse scenarios on that specific line. This reveals if a favorite's dominance is fragile. If a team is -1.5 favorites but rated 'Volatile', they may still win, but the margin required to cover the spread is at risk if they concede early.
How It Works
1. **Stability Score:** We simulate the probability of covering the line if the team concedes early (15') or trails at halftime. The score is the worst-case ratio compared to the pre-match probability. 2. **Entropy Check:** We verify if the match has high intrinsic goal potential (High Tempo) using a league-adjusted entropy lift.
The Signals
🟢 **Stable:** Probability holds firm even under stress. 🟡 **Sensitive:** Probability degrades under stress; goals pivot recommended. 🔴 **Volatile:** Probability collapses under early adversity. Avoid the handicap.
Basketball Methodology
🎯 Four-Layer Stack
Season baseline with prior merge, SRS-lite opponent adjustment, recency blend (last K games), and venue/splits adjustments. Optional enrichment when lineup/injury data is available.
📊 6 Core Markets
Win/Loss, Total Points O/U, First/Second Half O/U, and Team Totals—no draws in basketball.
📈 Full Transparency
Technical Notes show each layer's contribution. Data Sources disclose APIs, sample sizes, and caveats. AI Analysis provides insights on-demand.