P3's cluster model uses biomechanical data from over 1,000 NBA players to identify seven distinct athletic archetypes — each with different physical signatures, different ceilings, and different implications for how a career unfolds.
The NBA draft combine measures jump height, sprint speed, lane agility, and wingspan. These numbers get published, debated, and eventually filed away. What they do not do is tell you how an athlete actually moves — the mechanical strategy underneath the output, the joint-level signatures that determine whether a player's physical tools translate to sustained NBA performance or disappear within two seasons.
P3 has assessed more than 1,000 NBA players using 3D motion capture and force plates. That dataset — the largest elite athlete biomechanics database in the world — contains approximately 500 variables per assessment. The question was whether the data would reveal natural groupings: distinct athletic profiles that organize how NBA athletes move, produce force, and distribute mechanical load across their bodies.
It does. Using k-means clustering across ten biomechanical and anthropometric variables, P3 identified seven distinct athlete archetypes in the NBA population. The clusters are stable across repeated analyses and account for the vast majority of variability in the dataset. Each archetype carries different physical signatures, different positional distributions, and — critically — different implications for career trajectory.
The finding
Seven
archetypes
Traditional Bigs. Size is the dominant characteristic. Height and reach sit at or above the 90th percentile relative to the full NBA population. Vertical power output and lateral movement efficiency are limited. These athletes rely on physical dimensions to occupy space. Without an elite complementary skill — shooting range, passing vision, rim protection instincts — career lifespan in this cluster tends to be short. The modern NBA has compressed the window for athletes whose primary advantage is being large.
Bigs Plus. The physical profile teams want in a modern big. Height and reach remain in the upper tier — 80th and 70th percentile, respectively — but vertical plane output is substantially stronger and concentric force production is elevated. These athletes combine length with explosiveness.
Minus Perimeter. Lower-third height and reach. Limited lateral plane movement efficiency. Below-average vertical explosiveness. The cluster with the fewest standout physical qualities. Athletes who sustain careers in this group tend to possess extreme skill — elite shooting, elite passing, elite basketball IQ — that compensates for the absence of a clear physical advantage.
Kinematic Movers. This is where movement quality becomes the defining trait. Lateral movement skills are the dominant characteristic — specifically, hip mechanics during lateral acceleration tasks. Skater hip abduction and skater hip velocity emerge as key differentiators from the cluster analysis. These athletes may not test as the most explosive on traditional combine metrics, but they create separation through tempo, rhythm, and the efficiency of how they change direction. They tend to be guards and wings who play with a fluidity that is difficult to quantify with a stopwatch but immediately visible on film.
Force Movers. High force production numbers across the board. Strong concentric force, general explosiveness, solid traditional performance metrics. These athletes tend to test well on one-direction combine measures — vertical leap, approach jump, sprint speed. The gap is in change-of-direction efficiency. Raw power without the lateral movement quality to redirect it. In a league that demands multi-directional athleticism, force alone has not consistently translated to on-court impact at the level the combine numbers would suggest.
Specimens. Mid-range height and reach — roughly the 45th percentile — but elite explosive depth vertically and laterally. These athletes combine size with movement quality across both planes. The vertical delta is high. Lateral force production and movement efficiency are strong. This cluster tends to produce athletes with genuine versatility: big enough to play physical, explosive enough to play above the rim, and mechanically efficient enough to guard multiple positions.
Hyper Athletic Guards. The smallest athletes in the dataset — 17th percentile in height, 12th in reach — but their vertical and lateral movement numbers are exceptional. Super powerful relative to body mass, highly efficient movers in every direction. To survive at the NBA level at this size, everything has to work at an elite level. There is no room for a physical hole. The athletes who land here and sustain careers are the ones whose movement profiles are so outstanding that they overcome the structural disadvantage of their dimensions.
The pattern
What the clusters
reveal
The clustering is unsupervised — the algorithm finds the groupings without being told what to look for. It uses ten variables that span anthropometrics, vertical force production, lateral movement mechanics, and kinematic efficiency. Principal components analysis reduces those ten dimensions to two for visualization, but the clustering itself operates across the full dimensionality of the data.
Several patterns emerge. First, the clusters are stable. Repeated iterations with different random starting points produce the same core groupings, with variation only at the margins — edge cases where an athlete sits between two profiles. Most athletes stay in their cluster across multiple assessments over time.
Second, the clusters are not a hierarchy. There is no single best archetype. Traditional bigs have shorter average career spans in the current NBA, but that reflects how the league values certain physical profiles — not an inherent ranking of athleticism. A kinematic mover is not better than a specimen. They are different mechanical architectures with different strengths, different vulnerabilities, and different developmental pathways.
Third, and most importantly for front offices: two athletes with identical combine numbers can belong to entirely different clusters. The output metrics — how high, how fast — may match. The mechanical signatures underneath — how force is produced, how joints coordinate, how movement is organized — may be fundamentally different. Those mechanical differences have implications for durability, for defensive versatility, for how a player's game ages.
Why this changes evaluation
Traditional scouting combines the eye test with output metrics. Film shows you what a player does. Combine testing shows you how high and how fast. Neither shows you the movement architecture that produces those outcomes.
The cluster model gives teams a biomechanical framework for understanding what kind of athlete they are evaluating — not just how athletic, but in what way. A prospect who clusters with kinematic movers has a different development curve, different injury risk profile, and different positional ceiling than one who clusters with force movers, even if their vertical leap is identical.
For player development, the clusters also clarify what is trainable and what is structural. An athlete's cluster assignment is largely stable, but there is plasticity at the edges — particularly between kinematic movers, force movers, and specimens. Training interventions that shift an athlete's lateral mechanics or force production profile can, in some cases, move them toward a more favorable cluster. That shift is measurable, trackable, and directly connected to on-court movement quality.
The standard for understanding athleticism in basketball has been one-dimensional: how high, how fast, how big. The biomechanical reality is that elite athletes achieve their outputs through fundamentally different mechanical strategies. The cluster model makes those strategies visible — and actionable.
Methodology note
P3's athlete archetype model uses k-means clustering applied to biomechanical and anthropometric data collected from over 1,000 NBA players assessed using eight-camera 3D motion capture systems (220 fps) synchronized with force plate data (1000 Hz). The ten clustering variables include height, weight, reach, vertical plane outputs (delta vertical, concentric force, concentric rate of force development), and lateral plane mechanics (skater hip abduction, skater hip velocity, lateral force production, and lateral movement efficiency metrics). Principal components analysis is used for two-dimensional visualization. The seven-cluster solution was selected based on iterative stability analysis and interpretability. Cluster assignments are generated through an unsupervised process — no outcome labels are provided to the algorithm.
How P3 applies this: Talent Identification.
Related research: NBA players jump the same height in three completely different ways and Lateral force is what separates fast shufflers from slow ones.
More from P3: all P3 research. For teams and organizations: talk to P3. For high school athletes: the P3 College Prep Program.