Every look.
A little more context.
Inspect a program’s shot selection or an individual player’s attempts. Start with box-score-matched games, then examine the gaps in the evidence.
Loading the shooting lab…
The evidence behind the dots.
These are recorded field-goal attempts, not tracking data. Coordinates come from the source and may be approximate. Missing, placeholder, out-of-bounds and contradictory positions are excluded from the map while the attempts remain in shooting percentages. Long shots beyond half court also remain in the totals.
The default sample requires exact agreement with team FGA, FGM, 3PA and 3PM. Player samples must additionally match that player’s four box-score totals in the same game. Matching totals is a consistency check; it cannot prove that every event or location is correct. Unmatched games and players without an identity remain accounted for in coverage.
Shot types follow the publisher’s event labels. A layup is not a measured rim-distance bin. Field-goal points per attempt excludes free throws and is not offensive efficiency per possession. Split selection does not alter preseason forecasts. These historical affiliations do not establish a current roster.
2,915,731 total source events across 6,275 games; 6,300 completed games in the corresponding schedule. 700 accepted attempts lack a player ID; 14 potential attempts lack a valid team or event identity and are excluded. Counts cover all observed opponents, including programs outside the 366-team rating field.
Source: SportsDataverse’s bulk play-by-play release, CC BY 4.0. Normalization, coordinate checks and reconciliation by Silvermine. Publisher coordinate transformation. Model and source notebook →
Source receipt and calculation limits
SHA-256: 411d4ea11dba7d6688d717e26b3a6807640ae7f3b615b054a775e6f814d5c672
Basket-relative coordinates use the publisher’s origin (25, 0). Coordinate checks remove basket-center/default positions, values outside the court, three-point locations within 20 feet, at-basket shot types over 10 feet, and locations differing from a stated shot distance by more than four feet. These conservative checks do not correct or impute locations. Attempt values of zero are recovered only when the explicit source score-value field is two or three; that field describes attempt value even on misses.