How basketball analytics works

A basketball box score — points, rebounds, assists, shooting percentage — tells you what a player did but not how efficiently they did it, how much of the offense ran through them, or what happened on the 90% of possessions where they didn't touch the ball. Basketball analytics exists to answer those questions, and it has become central to how front offices build rosters, how coaches allocate minutes, and how broadcasters explain what viewers are watching.

The questions analytics is built to answer

Three questions come up constantly, in slightly different words, from a general manager, a coach and a fan: Is this player actually good, or just productive on a bad team? Should this lineup play more minutes together? And is a hot or cold shooting stretch a real change in ability, or noise that will fade? Analytics doesn't replace scouting or coaching judgment on any of these — it narrows the range of reasonable answers and flags when the box score is telling a misleading story.

The data it runs on

Three layers of data feed basketball analytics, and they differ in who collects them and how much context they carry.

• Box score data — points, shots made and attempted, rebounds, assists, turnovers, fouls — has existed for decades and is complete for essentially every professional and college game.
• Play-by-play and play-type data breaks a game down into possessions and tags what happened on each one: pick-and-roll, isolation, post-up, transition. Synergy Sports built its reputation on exactly this — analysts manually tag every possession by play type and link it to video, which is why "he shoots 42% out of pick-and-rolls" is a sentence a scout can actually check.
• Optical player tracking data comes from multi-camera systems installed in arenas that record every player's and the ball's position many times per second, without any device worn by the athlete. Second Spectrum, now part of Genius Sports, is a leading provider of this kind of tracking in professional basketball, and it's what makes metrics like speed, distance covered, and defensive closeout speed possible at all.

Video remains the connective tissue between the two: a coach or scout wants to see the possession behind the number, not just the number, which is why platforms built for tagging and clipping game film, such as Hudl, sit alongside pure stats products in most programs' toolkits.

Measuring scoring efficiency

Raw field-goal percentage undercounts three-point shooters, because a made three is worth 50% more than a made two but counts the same in the denominator. [[true-shooting-percentage|True shooting percentage]] corrects for this by weighting points against total shooting attempts, including free throws: TS% = points / (2 × (FGA + 0.44 × FTA)). It is the standard way to compare a high-volume three-point shooter to an efficient post scorer on a level footing.

[[usage-rate|Usage rate]] estimates the share of a team's possessions that end with a given player shooting, turning the ball over, or drawing shooting fouls while they are on the floor. It answers a different question than efficiency: not "how well did they score" but "how much of the offensive load did they carry." A player with modest efficiency and very high usage may still be more valuable than a hyper-efficient player who only operates in low-pressure situations — usage and efficiency have to be read together, not separately.

Measuring overall contribution

[[plus-minus|Plus-minus]] — the point differential while a player is on the court — is simple and intuitive but noisy: it is heavily influenced by which four teammates and which five opponents happen to share the floor at the same time, not just the player being measured. [[box-plus-minus|Box plus-minus]] tries to estimate a similar "overall value" number, but derives it statistically from box-score inputs (points, rebounds, assists, and so on) rather than from actual on-off point differentials, trading some real-world grounding for less noise.

[[player-efficiency-rating|Player efficiency rating]] (PER) compresses a full box score into one per-minute number, useful as a fast comparison across large numbers of players. Every composite rating in this family shares the same limitation: it is built almost entirely from data that is easy to count, which means defense — positioning, contesting shots without fouling, help rotations — is chronically under-measured relative to offense, because most of what a good defender does never shows up as a steal or a block.

How the work is done in practice

An analytics department typically combines the three data layers rather than picking one. A trade evaluation might start with box-plus-minus to screen candidates, move to play-type data from a provider like Synergy to check whether a target's production depended on a specific system that won't exist on the new team, and finish with tracking data and film to verify shot selection and defensive positioning the box score can't show. Coaches use a lighter-weight version of the same pipeline in-season: tracking-derived fatigue signals and play-type efficiency inform substitution patterns and opponent scouting, usually turned around within a day of the previous game using tagged, clipped video.

[[pythagorean-expectation|Pythagorean expectation]] — a formula that estimates a team's expected win total from its points scored and allowed, rather than its actual win-loss record — is used at the team level to judge whether a record overstates or understates true team strength, which matters for evaluating a coaching staff or setting expectations for the second half of a season.

Common mistakes and misreadings

• Treating a hot shooting stretch as a skill change. A player shooting well above their career three-point rate for a few weeks is very often regressing to the mean soon, not becoming a permanently better shooter — the sample is simply too small to tell the difference yet.
• Comparing usage rate without efficiency, or efficiency without usage. Each number alone tells a partial and sometimes misleading story about a player's value.
• Reading plus-minus over a small number of games as if it reflects a player's true impact — lineup effects and opponent quality dominate the number until the sample grows into the hundreds of possessions.
• Assuming a composite rating captures defense fairly. Most publicly available metrics still lean heavily on offensive box-score inputs; a defensive specialist can look mediocre on paper while being genuinely valuable on the floor.
• Ignoring play-type context when trading for or signing a player. A scorer who thrived in a system built around them may not replicate that production in a different offensive structure.

For a wider look at how teams combine tracking, video and injury-risk data across the business as a whole, see how sports teams use analytics, and for choosing a performance-data vendor directly, how to choose a sports performance analytics tool. Browse every tool in this category.