NBA 3-Point Case Study
In the 2014-15 season, NBA players made about 35% of their three-point attempts. What drove that number — and how would the answer change for a different player, a different defender, or a different moment in the game? Four tools let you explore.
Five Tools, One Dataset
Two simulators run the published make-rate account; a guided walkthrough opens the model up; an explorer and a model builder let anyone inspect, challenge, or extend the published case study.
All-or-Nothing Simulator
Pin every shot to one chosen condition per factor — what if every shot were taken with the defender 9+ feet away? See how the league-wide make rate shifts under the published model — or a player-specific refit when you filter to a single player.
Try itFine-Tuning Simulator
Same published model — or a player-specific refit — but shift the mix of shot conditions gradually: move 20% of tight-defense shots to open rather than moving every shot at once. Closer to how coaching adjustments actually play out.
Try itInside the Model
A guided walkthrough of how the published model turns one shot's six conditions into a predicted make probability — and how the full set of per-shot predictions collapses into the league make rate.
Try itShot Log Explorer
See how shots were distributed across conditions, which conditions moved together, and how one group of shots compared to another. Frequencies, correlations, and crosstabs.
Try itModel Builder
Refit the model with whichever shot conditions you choose. Add or remove predictors, fit on a single player, or let Auto-Build find the best-performing combination. See how the published account holds up across the league — and where individual players break the pattern.
Try itAbout the 2014-15 NBA Season
A Pivotal Season for the 3-Point Shot
The rise of the three
Teams attempted a then-record 22.4 three-pointers per game, up from 18.0 just five seasons earlier.
Splash Brothers era
Golden State won its first title in 40 years behind MVP Stephen Curry's record-breaking 286 made threes.
SportVU tracking
The NBA's SportVU cameras recorded shot-level data — defender distance, shot clock, dribbles, touch time — for the first three-quarters of the season.
Defender Distance
How close was the nearest defender? (0-3, 3-6, 6-9, or 9+ ft.)
Shot Distance
From the center of the basket (22-24 up to 26+ ft.)
Shot Clock
Late (under 4s) or open, plus touch time before the shot
Catch vs. Dribble
Did the shooter catch and shoot, or dribble first?
The Original Data
The NBA's SportVU optical tracking system (later Second Spectrum) recorded every shot in the first three-quarters of the 2014-15 regular season, producing the most granular public shot-level dataset the league has ever released. Most public analyses stopped at zone-by-zone make rates; this case study reopens the data — refit by player, change the defender-distance bins, or simulate what happens when you swap predictors.
Variables in the Published Model:
Beyond this case study
See how the same approach travels to other domains, or read how engagements are structured.