How we measure basketball
The methods behind our analysis and our public EuroLeague pages, starting with the paper we presented at the MIT Sloan Sports Analytics Conference 2026.
MIT Sloan Sports Analytics Conference 2026 · Basketball paper track
The paper: Scouting Anyone
Positions describe basketball less well every season. A stretch big who posts up now and then and defends the pick and roll does not belong to one role, yet the usual clustering methods (k-means, hierarchical clustering) put every player in exactly one group, and those groups gather around the average player instead of the real extremes of playing style.
Scouting Anyone: Probabilistic Player Archetypes for Any League uses Archetypal Analysis, which finds the extreme statistical profiles in a league, and Archetypoid Analysis, which ties each of those extremes to a real player. Every player-season-team is then described as a mix of archetypes that adds up to 100%, not as a single label.
The European model is trained on seven domestic first divisions (ACB, LNB, LBA, BBL, BSL, the Greek league and the ABA League) from 2014-15 to 2024-25, using only stats that exist for almost any league: height, rebounding, assist and turnover rates, steals, blocks, usage, free-throw rate, shooting splits and scoring volume. A player needs 300 minutes in a season to enter. The number of archetypes, eight, is where the players inside each archetype stop becoming more alike.
The output reads like a scout's sentence with numbers attached: Dylan Osetkowski's 2023-24 season at Unicaja is 57.0% High Usage Forward, 23.2% Role Guard, 15.2% 3PT Specialist and 4.5% Defensive Specialist.
Read the paper at MIT Sloan ↗From archetype to decision: the Howard case
The paper's case study retrains the model on 2024-25 EuroLeague data (249 player-seasons). Markus Howard came out 72% Shooter Specialist, one of the highest single-archetype concentrations in the league, and Baskonia were −7.0 in Net Rating with him on court and +3.0 with him off.
The archetypes explain the gap. 57% of his minutes came next to Traditional Centers, Mid-Range Specialists and High Volume Scorers, roles that compress spacing, while 3&D Wings, Stretch 4s and Rolling Bigs, the roles that make high-volume shooters work, shared only 22% of his time. Every pairing with his three most frequent teammates had a negative net rating. Measured against the eight Shooter Specialists who do lift their teams, the case points to two roster gaps: a rotation-level Stretch 4 and more 3&D wing minutes.
The player was not the problem; the ecosystem around him was. That is the kind of answer the method is built to give.
What is a player archetype?
A player archetype is one of the extreme statistical profiles in a league: the purest version of a way of playing, such as a high-volume scorer or a 3-point specialist. In our model a player is not given a single label: each season is described as a mix of archetypes that adds up to 100%, fitted separately for each league.
The eight EuroLeague archetypes (2024-25)
- Shooter Specialist
- Elite perimeter shooter, off the ball: a high share of threes, many spot-ups, little self-creation.
- 3&D Wing
- Two-way perimeter player: moderate three-point volume, a high steal rate, low creation and defensive versatility.
- Lead Guard
- Pass-first facilitator: a very high assist rate, a high turnover rate, low scoring volume and heavy pick-and-roll handling.
- Mid-Range Specialist
- Interior slasher with a midrange game: efficient inside the arc, few threes, pull-up oriented.
- Traditional Center
- Interior finisher and rim protector: the tallest, with high block and offensive rebound rates, post-ups and putbacks, and few threes.
- High Volume Scorer
- Multi-level scoring threat: high usage, high points per 100 possessions and a balanced shot profile.
- Stretch 4
- Big who spaces the floor: corner threes and pick-and-pop at moderate height.
- Rolling Big
- Vertical threat: roll man and putbacks, strong finishing at the rim, lobs, almost no perimeter shooting.
Each league gets its own taxonomy, fitted on its own data, so the labels change from one competition to another.
The method behind each tool
-
Archetypes
The paper's method, run per league and per season. Similarity between two players is the distance between their archetype mixes, which is how we find who plays like the player a club is about to lose.
How it works → -
Player Projection
Projects a player's production per 40 minutes into a destination league, adjusted for role and minutes, with a 10th to 90th percentile range and a confidence label for each stat. The range matters as much as the number: it says how far the projection can be trusted.
How it works → -
Expected Points
Every shot gets an expected value (xPPS) from its location, distance and shot type. Set against the points actually scored, it separates shot quality from shot making. Zones with fewer than 20 shots and play types with fewer than 15 plays are flagged as low confidence.
How it works → -
Lineup On/Off
Ratings for every player, duo and trio with them on court and off, counted possession by possession from the play-by-play. Clutch time, garbage time, home or away and the last N games are filters on the same data.
How it works → -
Clutch Data Score
Efficient offensive production plus defensive points saved from on/off, on one scale and cumulative over the season. It orders our EuroLeague MVP race.
How it works →
Data and conventions
- Possessions
- Counted, not estimated, wherever there is play-by-play. A possession ends in a shot attempt, a turnover or a trip to the free-throw line, and an offensive rebound keeps it alive. Each team keeps its own count.
- Garbage time
- Left out of efficiency metrics by default. A game enters garbage time with a margin of 20 points from minute 30, 15 from minute 35 or 10 from minute 38. The EuroLeague net rating applies this rule.
- Clutch time
- The last five minutes of the fourth quarter and any overtime, with the margin at five points or fewer.
- Samples
- A minimum of minutes or attempts on every model and ranking, and small samples flagged wherever they are shown.
- Each league on its own terms
- Models are fitted per league where the data allows, so a value means what it means in that competition.
- Ties in the standings
- We apply Article 19 of the EuroLeague Bylaws as written: the head-to-head only counts once the tied teams have all played each other twice, and until then the overall point difference decides. The EuroLeague standings explain every current tie.
Every term on this page, with its formula, is in the glossary →
The same methods, on your decision
For clubs we apply these methods to one concrete question: a signing, a rotation, a season that needs explaining. The first call is 30 minutes and confidential.
Request a confidential 30-minute call