Skip to main content
NWSL Notebook

Metrics Modeled

Expected points

xPts°Modeled

An estimate from a model we built. Useful, but not a fact — every modeled stat's page says where it's weak.

Definition

A scoreline-probability model turns each team's expected-goal total into probabilities of winning, drawing, and losing. Expected points weights those outcomes as three, one, and zero points. It is a process estimate, not a judgment about what a team deserved.

How it is computed

Estimate the compatible scoreline distribution from match-level xG, then calculate 3 × P(win) + 1 × P(draw).

How much is enough

A single match can be described, but team-strength comparisons need several matches from the same competition and model version.

Where it is weak

It inherits xG calibration limits and cannot represent red-card counterfactuals, unobserved tactical intent, or the value of protecting a game state.

What it refuses to measure

Expected points does not measure points deserved and should not replace the actual table.

What would prove this wrong

If held-out win, draw, and loss probabilities are materially less calibrated than the documented baseline across two consecutive evaluation windows, this model is voided.

Where it appears