Every NBA arena has cameras mounted in the catwalks that track every player and the ball at 25 frames per second. Machine learning in the NBA turns that tracking data into pregame scouting reports, real-time broadcast graphics, and the injury prevention models that decide whether a player sits out practice.
NBA Player Tracking Technology
From 2017 to 2023, Second Spectrum (now owned by Genius Sports) served as the NBA’s official optical tracking provider. Their cameras captured positional data for all ten players and the ball throughout every game.
In 2023, the NBA switched to Sony’s Hawk-Eye Innovations, which added 3D pose tracking. Hawk-Eye uses 12 cameras per arena to capture 29 points on each player’s body, not just a single location dot. That means the system tracks joint angles, body orientation, and limb position in three dimensions.
That adds up fast. Every second of game time generates positional data for 10 players, the ball, and 29 body points per player. A single game produces millions of data points.
Pattern Recognition and Basketball Analytics
Player coordinates alone don’t tell a coach much. The useful part is classifying what’s actually happening on the court.
Second Spectrum built pattern-recognition algorithms that could classify hundreds of complex basketball actions from tracking data alone. The system could identify whether a pick-and-roll happened, what variation it was, whether the ball handler accepted or rejected the screen, whether the screener popped or rolled, and how the defense responded. Before this existed, classifying those actions meant human analysts watching film. The classifier handles it automatically across every play of every game.
That capability changed pregame scouting. Coaches could pull up season-long data on how opponents ran specific plays and which defensive strategies worked best against them. The San Antonio Spurs’ analytics staff used this kind of analysis in the 2014 Finals, tracking how Miami Heat players’ shooting percentages changed based on who was guarding them and at what distances. The Spurs held LeBron James and Miami to barely 90 points per game.
Player Health and Load Management
The tracking data also feeds into injury prevention. Teams build models that monitor each player’s movement patterns, cumulative workload, and game-to-game changes in how they move, looking for early signs of fatigue or mechanical breakdown.
Research published in the American Journal of Sports Medicine evaluated machine learning models for predicting lower extremity muscle strains in NBA players using 20 years of performance data. The XGBoost model, an ensemble method that combines many decision trees, outperformed traditional logistic regression, using an 8-week sliding window of player activity to provide real-time injury risk snapshots.
The newer Hawk-Eye pose data makes this more precise. Instead of just knowing a player ran 2.3 miles during a game, teams can now see how their movement mechanics change as they fatigue — landing angles getting steeper, deceleration getting slower. One NBA team reported a 37% reduction in non-contact lower-body injuries over two seasons after implementing biomechanical analysis from tracking data.
Machine Learning in NBA Broadcasts
The same tracking infrastructure powers what viewers see on screen. The NBA’s alternate telecasts on League Pass use Second Spectrum’s augmentation technology to overlay real-time analytics directly onto the broadcast. Shot probability, defensive matchup quality, and play classification appear as graphics integrated into the live feed.
It’s the same machine learning pipeline used differently. The algorithms that classify a play as a pick-and-roll for a coaching staff also generate the on-screen graphic that tells a viewer “this was a high pick-and-roll, and historically the Warriors convert at 52% from this action.”
One Data Pipeline, Multiple Machine Learning Applications
The NBA collects the data once, then builds different models for different audiences on top of it. On the team side, tracking data goes into scouting, game planning, and injury prevention. On the fan side, it creates a broadcast experience that didn’t exist ten years ago.
You see the same setup in healthcare, manufacturing, and most other industries running machine learning in production: one central data pipeline, many models, many stakeholders.
For more on how the NBA’s analytics ecosystem developed, see this Yahoo Sports deep dive into why AI has become the NBA’s biggest competitive secret.
