Skip to main content

Prediction: Dependency

LineupAPI publishes predictions up to 17 days in advance. At that horizon, the biggest unknown is simple:

Will the player actually start the match?

Because playing time drives almost everything that happens in a game, most player predictions are to be interpreted:

“What are the chances if the player starts?”

What the lineups group means

lineups models three match-involvement outcomes:

  • lineups.starter — chance the player is in the starting 11
  • lineups.benchAndPlay — chance the player is on the bench and subbed in during the match
  • lineups.notPlay — chance the player does not play at all
    (out of squad, or unused substitute)

Think of this as: starts vs. comes on vs. no minutes.

The dependency principle

Important

Most other groups show only probabilities that hold, if the player starts the match. This applies to:

  • minutes
  • goals
  • assists
  • cards
  • corners_taken
  • fouls_won
  • fouls_conceded

Why we publish it this way

A player’s natural chance of goals, assists, cards, fouls, etc. is mostly a function of:

  • whether they start, and
  • how many minutes they are likely to play.

A bench-only appearance usually means fewer minutes and fewer opportunities.

Instead of forcing every group to “guess” playing-time scenarios internally, LineupAPI exposes the lineup signals (starter, benchAndPlay, notPlay) and the production signals (goals/assists/cards/etc.) so you can combine them into the exact product behavior you want.

How to use this correctly in your product

Option A — Build an “effective probability” (simple weighting)

If you want a single number that accounts for playing likelihood, you can combine starter and benchAndPlay into a rough play factor.

Example heuristic (tune the bench weight to your use case):

  • Play factor
    playFactor = lineups.starter + 0.2 * lineups.benchAndPlay

  • Adjusted threshold probability (example: Goals > 0.5)
    adjusted = playFactor * goals.over_0.5

This is not an official model output — it’s a practical UI/ranking trick many products use.

Option B — Show a small context note next to stats

When you display goals, assists, or minutes, add a short hint like:

“Projections should be interpreted with starter probability.”

This prevents “why is this high/low?” confusion when lineup likelihood is the real driver.

Option C — Filter or de-emphasize low-opportunity players

Common UX patterns:

  • Show only players above a starter threshold (e.g. lineups.starter >= 0.40)
  • Grey out players with high lineups.notPlay
  • Sort by lineups.starter by default