Expected points in basketball

What it does

• Shot quality, separated from shot making.

  Every shot carries an expected value, xPPS: what the model predicts the shot is worth, based on its location, distance and shot type. Set against the points actually scored, it shows whether a team is creating good looks, converting them, or both.

• Zone by zone, play by play.

  The court split into ten zones, and twelve play types grouped as pick-based, movement-based, self-creation and in flow, each with expected against actual.

• Any set of players.

  Choose who is on the floor for each side and compare: a pair on against the same pair off, or one lineup against another. The filter counts every shot taken while those players were on court, not only their own.

• Small samples flagged.

  Zones with fewer than 20 shots and play types with fewer than 15 plays are marked as low confidence.

• The questions it answers.

  Are we getting bad shots, or good shots we are missing? What happens to our pick and roll when a given player sits? Who is taking the shots in each zone, and how well?

• Inside an engagement.

  Expected Points is one of the tools our analysts use in client work, above all in performance diagnostics and opponent scouting. What the club receives is the analysis and the argument behind it.

Glossary

xPPS

Expected points per shot. The model's prediction of what a shot is worth, based on its location, distance and shot type.

Actual PPS

Actual points per shot: total points scored divided by total shots.

Diff

Actual PPS minus xPPS. Positive means the team or player scored above expectation, negative means below.

On/Off

Filters shots by whether the selected players were on court when the shot was taken, not by who took it.

*

Low confidence: fewer than 20 shots in a zone, or fewer than 15 plays in a play type.