Most football "prediction" apps work on hype. Pitch IQ works on math. Here's exactly how our engine generates the 11 markets you see for every match.

Layer 1: Poisson Distribution on Expected Goals (xG)

Every team has an attack strength and defense strength based on goals scored and conceded across recent matches. We compute expected goals (xG) for each side, then run a Poisson distribution to get the probability of every possible score (0-0, 1-0, 2-1, etc.). Sum those up and you have win/draw/loss probabilities.

Layer 2: Recent Form Weighting

The last 5-10 matches matter more than season averages. We apply heavier weight to recent results — a team on a hot streak gets a small boost, a team in a slump gets a small drag. This catches form swings the season average misses.

Layer 3: Head-to-Head History

Some matchups defy form. Atletico Madrid vs Barcelona doesn't go to script even when Barcelona dominate La Liga. We pull real H2H data from API-Football and blend it into our base probabilities.

Layer 4: Bookmaker Odds Blending

Bookmakers know things models don't — injuries, suspensions, motivation. When odds are available, we blend our model 60% with market odds 40%. You'll see a ✨ "Model + bookmaker odds blended" tag when this happens.

What Comes Out

11 separate market predictions per match: Match Result, BTTS, O/U Goals, Asian Handicap, Double Chance, Correct Score, HT/FT, Draw No Bet, Corners, Cards, Most Goals Half. Each with its own probability.

The Honest Part

No model is perfect. Football is genuinely random at the margins. Our Match Result accuracy on competitive matches sits around 55-65% — what professional models hit. We publish wins AND losses on the Upgrade page so you can verify before paying.