Back to Documentation
Model Documentation

Soccer Models

Deep-dive into the methodology, edge detection, and operational details for our EPL prediction models powered by expected goals (xG) analytics.

Available Models

Soccer Model Overview

Our soccer models are built on expected goals (xG) analytics, leveraging historical shot quality data to project match outcomes. Key principles include:

  • 1.xG-based ratings - Team attack and defense ratings derived from expected goals for and against, providing a more stable signal than actual goals.
  • 2.Poisson/Skellam distributions - Goals follow Poisson distributions. We use this to calculate precise over/under probabilities and margin distributions.
  • 3.Time-weighted ratings - Recent matches weighted more heavily than early-season games to capture current form while maintaining sample size stability.
  • 4.EV-based filtering - Every pick must have positive expected value (EV) based on our model probability versus the implied book probability.