| Bourbousson
et al. (2010a) |
To analyze
spatiotemporal coordination between basketball players during match play
using a dynamical systems approach. |
Data were
drawn from six game sequences in a 2008 French men's professional match,
including 10 players, 20 intra-team pairs, and 25 inter-team pairs. |
Relative
Phase Analysis; Coordination Dynamics. |
Dyadic
relative phase: the phase relationship between two players’ longitudinal
or lateral displacement trajectories, used to quantify in-phase, anti-phase,
or transitional coordination between player dyads. |
Player
dyads showed strong in-phase relations in the longitudinal direction, especially
for matched player–opponent dyads, indicating movement coupling shaped
by game demands. |
This metric
describes dyadic coordination, but it does not specify the tactical purpose
of the coupling or whether the observed relation was beneficial for either
team. |
Based on
six sequences from one professional match; player location was estimated
from video tracking, and tactical context was not fully linked to the observed
phase patterns. |
| Bourbousson
et al. (2010b) |
To examine
spatiotemporal coordination between two teams during match play by analyzing
team-level movement dynamics within a dynamical systems framework. |
Data were
drawn from one professional men’s basketball match, including six
selected game sequences involving two teams (10 players). |
Relative
Phase Analysis. |
Team spatial
center and stretch index: the geometric center of each team and the mean
distance of players from that center, used to quantify team displacement,
expansion, and contraction. |
The two
teams showed strong longitudinal in-phase coordination, while stretch-index
patterns suggested reciprocal expansion and contraction between the attacking
and defending teams. |
These indicators
describe team movement and spacing, but they should not be read as direct
evidence of tactical effectiveness without information on possession phase
and action context. |
Based on
one professional match and six selected sequences; team-level averages may
mask individual roles and specific tactical actions. |
| Shortridge
et al. (2014) |
To quantify
spatial variation in basketball shooting efficiency by developing spatially
explicit metrics that account for shot location and league-wide shooting
tendencies. |
All field-goal
attempts from the 2011–2012 NBA regular season, including more than
140,000 shots taken by players with at least 250 shot attempts. |
Empirical
Bayes Spatial Smoothing. |
Spatial
relative field-goal efficiency: empirical Bayes-smoothed shooting efficiency
at court-specific locations, used to estimate location-adjusted shooting
performance. |
Spatially
explicit metrics differentiated players’ location-specific shooting
ability and showed that players with similar overall shooting percentages
may differ across court areas. |
These metrics
describe where shooting efficiency is higher or lower, but they do not show
how the shot location was created within the possession. |
Based on
shot-location data from one NBA season; defensive pressure, passing sequence,
and possession context were not directly included. |
| Esteves et
al. (2015) |
To examine
collective spatial distribution patterns in basketball in relation to offensive
and defensive performance outcomes. |
Three U14
competitive basketball games, including 10 selected offensive sequences. |
Multivariate
Analysis of Variance; Mixed-design ANOVA. |
Court-area
occupation ratio: the number of attackers, the number of defenders, and
the attacker-to-defender ratio within seven predefined court areas during
offensive sequences. |
Spatial
distribution differed by court area; defensive numerical superiority near
the basket was associated with successful offensive outcomes. |
These measures
describe local numerical distribution, but they do not identify the tactical
actions that created the numerical pattern. |
Exploratory
analysis of three U14 games and 10 selected sequences; findings are limited
by small sample size and youth competition context. |
| Esteves et
al. (2016) |
To examine
how space occupation near the basket influences collective behaviors and
performance outcomes in youth basketball. |
Ten competitive
games involving 13 U14 teams. |
SMD; Cross-
correlation; Multinomial Logistic Regression. |
Local shot-space
indicators: the number of attackers, the number of defenders, attacker–defender
interpersonal distance, and shooter-to- basket distance measured around
shot attempts. |
Larger
attacker–defender distance and shorter distance to the basket increased
the likelihood of a converted shot; defensive numerical overload near the
scoring target was also associated with offensive success. |
These indicators
describe local numerical and distance constraints near the basket, but they
do not identify the specific offensive action that created the advantage. |
Youth sample
only; selected sequences from U14 games; two- dimensional video digitization;
and limited generalizability to senior or professional basketball. |
| Franks et
al. (2015) |
To quantify
the independent influence of NBA defenders on opponents’ shot selection
and shooting percentages across court regions, and to describe spatially
encapsulated individual defensive ability. |
Approximately
115,000 half-court offensive plays lasting at least 5 seconds from the 2013–14
NBA regular season. |
Log-Gaussian
Cox Process; Hidden Markov Model. |
Defensive
shot frequency effect and defensive shot efficiency effect: model-based
estimates of how a defender alters opponent shot frequency and shot success
across court regions. |
Defenders
influenced shot frequency and shotefficiency in different ways, allowing
defensive impact to be mapped across court regions. |
These model-derived
effects depend on inferred defensive assignments and should be interpreted
within the team defensive context, rather than as fully isolated individual
defensive ability. |
Individual
defensive effects could not be fully separated from team defensive strategy
and teammate support; results were based on one NBA season of optical tracking
data. |
| Sampaio et
al. (2014) |
To identify
game performance profiles using NBA player-tracking and technical statistics,
and to distinguish key performance characteristics of All-Star and non-All-Star
players through clustering and discriminant analysis. |
Tracking
and non- tracking data from 548 players across 1,230 NBA regular- season
games in the 2013–14 season. |
Linear
Discriminant Analysis; k-means Cluster Analysis. |
Player-tracking
performance profile indicators: NBA tracking and non-tracking variables
describing shooting type, touches, passing, rebounding, speed–distance
activity, and defensive actions. |
All-Star
players were mainly distinguished by elbow touches, defensive rebounds,
close touches, close points, pull-up points, and lower defensive speed;
cluster analysis identified scoring, passing, defensive, and all-around
profiles. |
These variables
describe player performance profiles across a season, but they do not show
the possession-level tactical process that produced each action. |
All-Star
grouping was based on media selection, full-court video was unavailable
to verify tracking data, and the cluster analysis used only complete cases. |
| Skinner et
al. (2015) |
To estimate
player-skill parameters and lineup- interaction patterns, and to use them
to predict team offensive efficiency and assess how player skills contributed
to team performance. |
Simulated
data for 5,000 possessions, together with 780 offensive sequences from the
2011 NBA playoff game between the Grizzlies and the Thunder. |
Markov
Chain Model. |
Offensive
network transition model: player-skill and lineup- interaction parameters
used to represent possessions as transitions through offensive states and
to predict lineup performance. |
The model
inferred player skills from simulated data and showed that limited real-game
data could describe how a player interacted with different five-man lineups. |
The model
describes offensive structure under simplified network assumptions, but
it does not directly identify observed tactical intent or specific play
design. |
Player-tracking
data were not publicly available; the main validation used simulated data;
and the real-game application was based on a limited playoff sample. |
| Cervone et
al. (2016) |
To construct
a multiresolution stochastic process model to simulate changes in player
movement during a possession and predict the possession outcome. |
Tracking
data from 461 players in the 2013–14 NBA season. |
Multiresolution
Stochastic Process Modeling; Markov Chain Modeling; Bayesian Inference;
Maximum Likelihood Estimation. |
Expected
possession value: model-estimated expected points remaining in a possession,
derived from multiresolution stochastic modeling of player movement and
discrete game events. |
The model
showed that offensive value could be updated continuously during a possession
and could reveal how player decisions and spatial strategies changed expected
outcomes. |
EPV describes
model- estimated possession value, but its tactical meaning depends on how
the model defines states, transitions, and relevant spatial information. |
High computational
complexity; reliance on high-quality optical tracking data; and model assumptions
about possession states, transition kernels, and event segmentation. |
| Santana et
al. (2019) |
To analyze
basketball offensive structure through concatenated space- creation dynamics
and to characterize team tactical patterns. |
Four NBA
teams from the 2013–2014 season, including multiple matches and offensive
possessions. |
Chi-square
Analysis. |
Space-creation
dynamics concatenation classes: isolated, independent concatenated, and
dependent concatenated offensive tactical actions coded within ball possessions. |
The framework
differentiated team offensive structures; the Spurs showed longer sequences
with more dependent concatenations, whereas other teams used shorter or
more direct patterns. |
The metric
describes the sequence structure of offensive tactics, but it depends on
manual coding and does not directly quantify spatial position or defender
response. |
Limited
to four NBA teams; observational match analysis; reliance on expert-defined
coding categories; and no direct use of tracking-derived spatial coordinates. |
| Manisera
et al. (2019) |
To use
tracking-data visualization and clustering analysis to reveal players’
spatial distribution patterns and inform tactical optimization. |
133,662
data points from a 2016 friendly match in the Italian C-Gold League. |
k-means
Cluster Analysis. |
Spatial
phase clusters and transition probabilities: clusters of homogeneous player-spacing
configurations and transition matrices describing movement between game
phases. |
Cluster
analysis identified different offensive and defensive game phases and showed
how tracking data could support visualization and phase-based interpretation
of team movement. |
These clusters
describe recurring spatial configurations, but their tactical meaning depends
on how phases are labeled and linked to possession context. |
Single
friendly match from the Italian C-Gold League; home-team data only; device-based
tracking; and no opponent or ball trajectory included in the final modeling. |
| Daly-Grafstein
and Bornn (2021) |
To analyze
NBA shot trajectories to examine the impact of defense on shooting accuracy
and to assess perimeter defenders’ defensive ability and shooters’
resilience under defensive pressure. |
Data from
50,916 three-point shot trajectories in the 2014–15 NBA season. |
Bayesian
Regression; logistic Regression. |
Shot-trajectory
factors and trajectory-based shot-make probability: shot depth, left-right
deviation, and entry angle derived from in-game ball trajectories and used
to estimate shot-success probability. |
Contested
three-point shots showed greater variance in depth and left-right accuracy;
trajectory-based metrics provided more stable estimates of perimeter defensive
impact than opponent field-goal percentage. |
These metrics
explain how contests affect shot trajectories, but they do not fully identify
the defensive scheme, closeout timing, or help-defense context behind the
contest. |
Restricted
to NBA three-point shots; dependent on SportVU trajectory quality and modeled
shot paths; and focused on perimeter defense rather than all defensive actions. |
| Santos-Fernandez
et al. (2022) |
To estimate
the intrinsic dimension (ID) of high- resolution tracking data and reveal
differences in movement complexity and behavioral structure at the player
and team levels. |
Tracking
data from 15 randomly selected games in the 2015–16 NBA season. |
Intrinsic
Dimension Estimation; Bayesian Linear Models;Non- parametric Tests. |
Intrinsic
dimension: a Bayesian mixture-model estimate of the latent dimensionality
of player-tracking data, used to quantify movement complexity and dependence
in offensive sequences and shot-chart patterns. |
ID values
increased during phases such as creating space for passing and shooting,
declined near the end of plays, and were higher in game-winning or closer-margin
contexts. |
ID captures
movement complexity and unpredictability, but it does not specify which
tactical action or player decision produced the observed complexity. |
Fixed number
of mixture components; temporal dependence not fully modeled; no measurement-noise
component; and need for validation in larger samples. |
| Supola et
al. (2022) |
To analyze
the characteristics of secondary assists and their impact on scoring efficiency. |
Tracking
data from the first half of the 2015–16 NBA season, covering 531
games. |
Logistic
Regression; Linear Regression. |
Secondary-assist
indicators: potential secondary assists, receiver openness, modeled Expected
Points, and Average Points per Possession for shot opportunities created
through pre-assist passing sequences. |
Secondary
assists were associated with more open shots, higher expected scoring value,
and particularly productive corner-three opportunities. |
These indicators
describe the value of pre-assist ball movement, but they do not fully separate
designed play execution from defensive breakdown or late-possession adaptation. |
Reliance
on public SportVU and event records; exclusion of some foul-related outcomes
from the model; and systematic imprecision in modeled Expected Points for
some shooting areas. |
| Barron et
al. (2025) |
To use
density-functional fluctuation theory to describe NBA player-position distributions
and interaction patterns during games, and to identify differences in player
roles and tactical-structure features. |
Player-
tracking data from the 2022–23 NBA season through January 20. |
Density-
Functional Fluctuation Theory; Maximum Likelihood Estimation. |
Density-functional
fluctuation theory indicators: player-density fields, location-preference
parameters, player–player interaction terms, and player- gravity
estimates inferred from NBA tracking data. |
DFFT predicted
player locations, produced a team-position-based metric related to play
outcomes, identified defensive positioning tendencies, and quantified offensive
player gravity while accounting for teammate positioning. |
These indicators
describe spatial preference and interaction structure, but their tactical
meaning depends on the density representation and does not directly encode
play calls or defensive assignments. |
Player
influence was represented through two-dimensional Gaussian density fields;
player-specific analysis was limited to selected high-minute players; and
physical attributes such as height or leaping ability were not directly
included in the model. |
| Jiao et al.
(2025) |
To identify
spatiotemporal factors distinguishing successful and unsuccessful possessions
using large-scale NBA tracking data. |
Tracking
data from 632 NBA games in the 2015–16 NBA season. |
Independent
t-test; Mann– Whitney U test. |
Possession-level
spatiotemporal indicators: ball kinematics, offensive and defensive team
spatial structure, shooter-specific spatial variables, and contextual variables
calculated for set-play possessions. |
Successful
shot outcomes were mainly distinguished by shooter-related spatial variables,
including shorter shot distance, larger defender-related shot angle, and
greater separation from the nearest defender. |
These indicators
identify spatial differences between made and missed shots, but they do
not on their own explain the tactical process that created the shooting
condition. |
Observational
design; single NBA season; no player-specific contribution analysis; and
limited integration of technical, physical, and tactical variables beyond
the tracking data. |