How OddsIntel Works
What the model does, what each tier unlocks, and why you can trust the track record.
The Model in One Paragraph
Every morning the model blends a Poisson regression (goals as a statistical process), an XGBoost ensemble (36+ features: form, ELO, H2H, lineups, injuries, rest days), and a third-party prediction cross-check. It then collects 58 signals per match — market movement, team strength, news, context, live data. The final probability is calibrated weekly against actual results so it doesn't drift. Full technical methodology →
From Prediction to Value Bet
A prediction tells you who is most likely to win. A value bet tells you where the bookmakers have underpriced that probability. These are different things.
Step 1 — Free tier
62%
Model says Home wins with 62% probability
Step 2 — Pro tier
1.85 odds
Bookmaker implies 54% probability (1/1.85). You see the full spread across 13 bookmakers.
Step 3 — Elite tier
+8% edge
Model beats market by 8 pp — a value bet. Kelly stake calculated automatically.
What Each Tier Includes
Symbols & Icons Guide
Here's what every symbol on the matches page means at a glance.
Data Grade
Match Pulse
Most matches (~80%) are "Routine" — no special indicator shown. The pulse only flags the ~15-20% of matches with unusual activity.
League Header
Predicted Score
Odds & Movement
Form Strip
Signal Teasers
Match Status
Common Questions
What's the difference between a prediction and a value bet?
A prediction says 'Home is most likely to win at 62%'. A value bet says 'Home is 62% likely but the bookmakers are only pricing it at 54%, so you have an 8% edge'. You can predict correctly and still lose money if you consistently bet at bad odds. Value betting solves that.
What does edge % mean?
Edge = model probability minus bookmaker's implied probability. +8% means our model thinks the true probability is 8 percentage points higher than what the odds imply. Finding positive edge consistently is what produces profitable betting over time.
How reliable are the track record numbers?
Every settled prediction is logged automatically when the pipeline runs — no manual curation, no removing losses. You can verify every row at /performance.
Ready to see today's predictions?
Free forever. No credit card required.
Want the full technical detail? Read the model methodology →