Title: Scouting Anyone: Probabilistic Player Archetypes for Any League
Author: Marlin Myrte, Sebastian Buzzalino, Thierry Aymerich
Institution: 2026 MIT Sloan Analytics Conference
Publication Date: March 7, 2026
The Problem: Addressing the "Average" Player Inefficiency
Modern basketball decision-making still relies heavily on rigid positions, subjective role labels, or hard clustering models that force players into a single category. This creates inefficiencies in scouting, lineup construction, and cross-league evaluation—especially when roles are fluid, hybrid, and context-dependent. Traditional clustering methods also tend to identify “average” players rather than the true stylistic extremes that coaches and scouts think in terms of. As a result, teams frequently misdiagnose player fit, overpay for redundant skill sets, or undervalue players who could thrive in a different ecosystem.
Methodology: From Abstract Clusters to Real-World Prototypes
The researchers utilize Archetypal Analysis (AA) and Archetypoid Analysis (ADA) to solve these issues.
Archetypal Analysis (AA): Think of this like a color wheel. Just as any color can be described as a mix of "pure" red, blue, and yellow, AA describes every player as a percentage of "pure" stylistic extremes. A player isn't just a "Center"; they might be 60% "Rim Protector" and 40% "Stretch Big".
Archetypoid Analysis (ADA): While AA creates theoretical "perfect" extremes, ADA anchors these roles to real-world benchmark players. Instead of an abstract statistical average, a scout can describe a prospect as "40% stylistically similar to Nikola Vucevic and 35% to Mitchell Robinson".
By applying these models to simple box-score data and advanced play-type metrics, the framework remains accessible to teams at any budget level.
Why it Matters: ROI through Stylistic Alignment
For GMs and coaches, this framework directly improves surplus value identification and risk management. It enables clearer separation between player talent and player fit, reducing costly acquisition mistakes driven by surface-level production. Scouts can search globally for stylistic matches rather than positional lookalikes, opening undervalued markets. Coaches gain a diagnostic tool to explain why certain lineups fail despite “good players,” identifying spacing, creation, or defensive redundancies at a structural level.
Crucially, probabilistic roles align with how basketball actually works: players shift responsibilities based on teammates and schemes. This allows front offices to project how a player’s value will change after a trade, signing, or coaching change—turning context dependency from a blind spot into a controllable variable with measurable ROI.
ACTIONABLE TAKEAWAYS
Audit Your Roster Ecosystem: Map archetype distributions for your top lineups to identify redundancy, spacing compression, or missing complementary roles.
Upgrade Scouting Filters: Replace position-based shortlists with archetype similarity searches across leagues to uncover undervalued fits.
Stress-Test Acquisitions: Before trades or signings, simulate how a player’s archetype mix interacts with existing core players to forecast on/off impact.
Draft Optimization: Rather than drafting for "best player available," target specific archetype percentages (e.g., a "Stretch 4" with at least 50% "Pick-and-Pop" concentration) to maximize the efficiency of existing high-usage scorers.
Research Paper Link: Scouting Anyone: Probabilistic Player Archetypes for Any League