Skip to main content

Prediction: Outputs

Probabilities

Many fields are probabilities from 0 to 100 (%).

Examples:

  • lineups.starter = 77 → 77% chance the player starts
  • lineups.benchAndPlay = 22.84 → 22.84% chance the player comes on from bench as a substitute
  • lineups.notPlay = 0.16 → 0.16% chance the player neither starts nor comes off the bench as a substitute

Tip

  • For the lineups prediction group, these three outcomes are designed to cover all scenarios and sum up to 100%.

Threshold probabilities (over_0.5, over_1.5, …)

Groups like goals, assists, corners_taken, fouls_won, and fouls_conceded use threshold probabilities.

Example (goals):

  • goals.over_0.5 = chance of 1+ goals
  • goals.over_1.5 = chance of 2+ goals
  • goals.over_2.5 = chance of 3+ goals

How to read it:

  • These are threshold probabilities, not expected goals.
  • The chance of exactly 0 goals is 100% - goals.over_0.5.
  • The chance of under 1.5 goals is 100% - goals.over_1.5.
  • Higher thresholds must be smaller (or equal): over_2.5 ≤ over_1.5 ≤ over_0.5

About the models

LineupAPI’s prediction models are developed and maintained by the founder (graduate Data Scientist with 5+ years of football analytics experience).
They use a mix of state-of-the-art machine learning techniques and are designed to output well-calibrated probabilities (so values like 70% behave like ~70% over many similar cases).
For model details, versions, and training metadata, see: /predictions/metadata.