Principal component analysis to understand basketball performance: A systematic review of key performance indicators,workload monitoring and physical fitness

Abstract

Basketball performance analysis now generates large datasets containing multiple physical and technical-tactical variables from wearable sensors and tracking systems. However, variable intercorrelation often obscures meaningful patterns, complicating decision-making in athlete development, training prescription, and tactical planning. Principal Component Analysis (PCA) has emerged as a widely applied dimensionality reduction technique, though its specific application within basketball research remains to be systematically examined. Therefore, this systematic review aimed to: (a) identify and synthesize PCA studies across three basketball performance domains (workload monitoring, key performance indicators, and physical fitness); (b) evaluate methodological quality and reporting standards; (c) characterize component structures; and (d) provide methodological recommendations. Following PRISMA guidelines, a comprehensive search of Web of Science, PubMed, SPORTDiscus, and Scopus was conducted, yielding 32 eligible studies. Overall study design quality was predominantly excellent (87.5% scoring >75% on MINORS). However, substantial heterogeneity existed in PCA-specific methodological practices: 68.8% of studies failed to report factorability assessment procedures and only 40.6% conducted sample adequacy testing. Consistent component structures emerged across domains. In workload monitoring, PC1 captured composite load profiles (accelerations, decelerations, distance) and PC2 reflected speed or well-being. In performance indicators, PC1 reflected offensive production and PC2 represented rebounding and defense. In physical fitness, PC1 captured aerobic or neuromuscular time-dependent qualities and PC2 reflected force-power attributes. These findings provide frameworks for identifying essential performance variables and reveal substantial methodological heterogeneity requiring standardized reporting of factorability testing, component retention, rotation procedures, and structural validation.

Keywords

Athlete profiling agility endurance performance indicators strength training load sport analytics

Introduction

Basketball is a complex team sport characterized by intermittent high-intensity actions, rapid offensive-defensive transitions, and multidimensional performance demands.^1,2 Technological developments in data acquisition (e.g., wearable sensors, optical tracking systems, and automated video analysis platforms) now allow the simultaneous recording of hundreds of variables during both training and competition,^3–5 fundamentally transforming how performance is planned, evaluated, and managed.⁶ However, the high dimensionality of these datasets and the intercorrelation among variables can obscure meaningful patterns, complicating decision-making in areas such as athlete development, training prescription, and tactical planning.^7–9 This challenge underscores the need for statistical methodologies capable of reducing data complexity while preserving essential information.

In this context, Principal Component Analysis (PCA) has been established as one of the most widely used techniques to reduce dimensionality in sport science.^10,11 PCA is an unsupervised statistical method that transforms correlated variables into uncorrelated components while retaining maximum variance.¹² This is particularly relevant in basketball, where numerous technical, tactical, physical, and physiological indicators coexist and are highly interrelated. PCA has been applied across multiple contexts including player profiling and performance prediction,^13,14 as well as examining injury-performance relationships and success-differentiating factors.^15,16

However, its correct application requires adherence to specific statistical assumptions such as factorability testing, component retention criteria, and rotation methods. Inadequate factorability assessment may extract artificial components, while arbitrary retention and rotation decisions can substantially alter component interpretations.^17,18 Despite these considerations, a recent systematic review identified that these decisions are frequently unreported in team sports research,¹⁹ limiting the comparability of findings and compromising the reproducibility of results.

Therefore, this systematic review aimed to identify and synthesize PCA applications in basketball performance research across three domains (key performance indicators, workload monitoring, and physical fitness assessment), evaluate methodological quality and reporting standards, characterize component structures, and provide evidence-based recommendations for future research and applied practice.

Methods

Study design

According to established classification frameworks for research designs in sport sciences, this study corresponds to a non-empirical systematic literature review design.²⁰ The study was designed and executed following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement²¹ and adhered to recommendations for conducting systematic reviews in sports science research.²² The review protocol was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO: CRD420251180383) to ensure methodological transparency and minimize reporting bias, and all procedures were conducted in full accordance with the registered protocol.

Search strategy

A comprehensive literature search was conducted across four major electronic databases: Web of Science (including Web of Science Core Collection, MEDLINE, Current Contents Connect, Derwent Innovations Index, KCI-Korean Journal Database, Russian Science Citation Index, and Scielo Citation Index), PubMed, SPORTDiscus, and Scopus. The search was executed on October 19, 2025, encompassing all records indexed in these databases from inception to the search date.

A systematic search strategy was developed based on the PICOS (Population, Intervention/Exposure, Comparison, Outcomes, Study design) framework to ensure comprehensive identification of relevant literature.²³ The search string combined specific terms using Boolean operators as follows:

(“principal component analysis” OR “PCA”) AND (“basketball”) AND (“performance” OR “key performance indicators” OR “game statistics” OR “workload” OR “training load” OR “internal load” OR “external load” OR “fitness” OR “physical fitness” OR “speed” OR “jump” OR “agility” OR “change of direction” OR “strength” OR “endurance”)

This search string was adapted as necessary to accommodate the specific requirements and syntax of each database while maintaining consistency in the search logic. Reference lists of all included articles were manually screened to identify additional relevant studies that may not have been captured through the electronic database search. Any disagreements regarding study inclusion were resolved by consensus between two investigators (C.D.G-C and D.M-T) and arbitration by a third investigator (J.R) when needed.

Inclusion and exclusion criteria

Following the initial database search, all retrieved records were exported to reference management software for duplicate removal and preliminary screening. The titles, authors, journal names, and publication dates of all identified articles were systematically recorded in a Microsoft Excel spreadsheet.

After removing duplicate records, the remaining articles were evaluated against predetermined eligibility criteria based on the PICOS framework (Table 1). Studies were included if they: (a) involved basketball players of any age, sex, or competitive level; (b) explicitly applied PCA methodology; (c) analyzed workload monitoring, key performance indicators, or physical fitness outcomes; and (d) used observational study designs. Studies were excluded if they lacked methodological detail, involved non-basketball populations, or were non-original research (reviews, editorials, abstracts).

Table 1.

Inclusion and exclusion criteria for study selection following PICOS strategy.

No.	Item	Inclusion criteria	Exclusion criteria
1	Population	Basketball players of any age, sex, competitive level, or playing position studied in any basketball setting (professional, collegiate, youth, training, competition, laboratory)	Studies involving participants with neurological disorders or severe musculoskeletal injuries; wheelchair basketball; mixed-sport studies without extractable basketball data; studies where basketball context cannot be clearly identified
2	Intervention/Exposure	Studies explicitly applying PCA to basketball data with reported methodological details (e.g., factorability, data suitability testing, retention criteria or rotation method)	Studies only mentioning PCA without reporting any methodological details (e.g., factorability, data suitability testing, retention criteria or rotation method); studies using factor analysis or other dimensionality reduction techniques without PCA; studies applying PCA to sports other than basketball.
3	Comparison	Not applicable	Not applicable
4	Outcomes	Studies analyzing key performance indicators (points, assists, rebounds, etc.), workload monitoring (internal/external load), or physical fitness (speed, jump, change of direction, strength, endurance) using PCA	Studies not reporting PCA results or interpretations; reviews, meta-analyses, editorials, commentaries, case reports
5	Study Design	Observational studies (cross-sectional, longitudinal, cohort, descriptive) that apply PCA methodology with no language restrictions	Randomized controlled trials, systematic reviews, meta-analyses, narrative reviews, editorials, letters, opinion pieces, abstracts without full text

Screening strategy and study selection

Following duplicate removal, two independent reviewers (C.D.G-C and D.M-T) conducted a two-stage screening process. Initially, titles and abstracts were evaluated against the eligibility criteria to identify potentially relevant studies. Subsequently, full-text articles of all potentially eligible studies were retrieved and assessed in detail to determine final inclusion. Throughout both screening stages, disagreements between reviewers were resolved through discussion or, when necessary, consultation with a third reviewer (J.R) to reach consensus.

Data extraction and analyzed variables

A standardized data extraction form was developed based on the Cochrane Consumers and Communication Review Group's template. Two reviewers (C.D.G-C and D.M-T) independently extracted data from all included studies, with cross-verification to ensure accuracy. Discrepancies were resolved through consensus discussion, with a third reviewer (J.R) consulted when necessary. Inter-rater reliability was quantified using Cohen's kappa coefficient, yielding almost perfect agreement (κ= 0.93).²⁴

An analytical framework was developed to classify the methodological application of PCA across studies, guided by established guidelines for PCA application, including recommendations on sampling adequacy, component retention, and rotation procedures.^17,18 This framework encompassed: (i) data preprocessing and factorability assessment, (ii) component retention decisions, and (iii) reporting and interpretability of results. The following information was extracted systematically: (1) study characteristics (author, year, country); (2) sample characteristics (sample size, sex, age, competitive level, setting); (3) study aim; (4) PCA methodology (retention loading criteria, sample adequacy criteria, sphericity criteria, factor loading thresholds, cross-loading management, rotation method); (5) PCA results (number of initial variables, number of factors extracted, percentage of variance explained, variables extracted per component); and (6) main findings.

Studies were categorized into three domains: (i) workload monitoring (internal/external load during training or competition), (ii) key performance indicators (game statistics and performance metrics), or (iii) physical fitness assessment (speed, jump, agility, change of direction, strength, endurance).

Quality assessment

The rigor of included studies was evaluated using the Methodological Index for Non-Randomized Studies (MINORS),²⁵ a validated instrument designed specifically for assessing study design and conduct in observational research. The MINORS instrument comprises twelve items, with the first eight applicable to all study designs and the remaining four applicable exclusively to comparative studies. Each criterion was scored as 0 (not reported), 1 (reported but inadequate), or 2 (reported and adequate), yielding maximum possible scores of 16 for non-comparative studies and 24 for comparative studies.

Quality appraisal was conducted independently by two reviewers (C.D.G-C and D.M-T), with inter-rater reliability quantified using Cohen's kappa coefficient.²⁴ Disagreements were resolved through discussion, with arbitration by a third reviewer (J.R) when consensus could not be achieved. Final quality scores were converted to percentages and categorized as: (1) low methodological quality (<50%), (2) good methodological quality (51–75%), or (3) excellent methodological quality (>75%), following established cut-offs for MINORS score interpretation.²⁵

Results

Study selection

After analyzing all databases (Web of Science: 150; PubMed: 40; Scopus: 93), 283 articles were initially identified. After removing duplicate records and excluding non-peer-reviewed publications and records outside the defined eligibility scope, 75 records remained for screening. Screening of titles and abstracts resulted in the exclusion of 23 articles that did not meet the eligibility criteria. The eligibility of the remaining 52 full-text articles was assessed, of which 20 were excluded for the following reasons: 7 articles did not report PCA methodology in sufficient detail, 2 articles did not focus on basketball populations, 3 articles were not original research (reviews, editorials, or conference abstracts), and 8 articles did not report results in the three specified domains. Finally, 32 studies met all inclusion criteria and were included in the synthesis of this systematic review^26–57 (Figure 1).

Figure 1.

PRISMA flowchart.

Quality assessment

The quality assessment of the included studies (n = 32) is presented in Table 2. Inter-rater reliability between the two independent reviewers was excellent, with a Cohen's kappa coefficient of 0.92 (95% CI: 0.85–0.97). Overall study design and conduct quality, as assessed by MINORS, was high, with scores ranging from 56.3% to 100%. Most studies (n = 25, 78.1%) were classified as excellent quality (>75%), six studies (18.8%) were classified as good quality (51–75%), and only one study fell within the good-to-moderate range at 56.3%. Across the three domains, workload monitoring studies demonstrated the highest average methodological quality (90.4%), followed by physical fitness studies (84.2%) and key performance indicators studies (82.1%).

Table 2.

Methodological quality assessment of the included studies.

Reference	1	2	3	4	5	6	7	8	9	10	11	12	Score	%
Workload monitoring
Salazar et al. (2020) ²⁶	2	1	2	2	2	2	2	0	2	2	2	2	21/24	87.5
Saucier et al. (2021) ²⁷	2	2	2	2	2	2	2	0	2	2	2	2	22/24	91.7
Zhang et al. (2026) ²⁸	2	2	2	2	2	2	2	0	-	-	-	-	14/16	87.5
García-Rubio et al. (2024) ²⁹	2	1	2	2	2	2	2	0	2	2	2	2	21/24	87.5
Ibañez et al. (2025) ³⁰	2	2	2	2	2	2	2	0	2	2	2	2	22/24	91.7
Svilar et al. (2018) ³¹	2	2	2	2	2	2	2	0	2	2	2	2	22/24	91.7
Rojas-Valverde et al. (2020) ³²	2	2	2	2	2	2	2	0	2	2	2	2	22/24	91.7
Stone et al. (2022) ³³	2	1	2	2	2	2	2	0	2	2	2	2	21/24	87.5
Zhang et al. (2025) ³⁴	2	2	2	2	2	2	2	2	-	-	-	-	16/16	100
Pino-Ortega et al. (2022) ³⁵	2	1	2	2	2	2	2	0	2	2	2	2	21/24	87.5
Key performance indicators
Oliveira & Newell (2024) ³⁶	2	2	1	2	2	2	2	2	-	-	-	-	15/16	93.8
Ke et al. (2024) ³⁷	2	1	1	2	2	2	2	0	-	-	-	-	12/16	75.0
da Silva & Rodrigues (2022) ³⁸	2	1	1	2	2	2	2	0	-	-	-	-	12/16	75.0
Doroshenko et al. (2020) ³⁹	2	2	1	2	2	2	2	0	2	2	2	2	21/24	87.5
Sampaio, et al., (2010) ⁴⁰	2	2	1	2	2	2	2	0	2	2	2	2	21/24	87.5
Contreras et al., (2025) ⁴¹	2	2	1	2	2	2	2	0	2	2	2	2	21/24	87.5
Péndola-Reinecke et al., (2025) ⁴²	2	2	1	2	2	2	2	0	2	2	2	2	21/24	87.5
Katris (2023) ⁴³	2	1	1	2	2	2	2	0	-	-	-	-	12/16	75.0
Yin (2014) ⁴⁴	2	0	1	2	2	1	1	0	-	-	-	-	9/16	56.3
Courel-Ibañez et al., (2025) ⁴⁵	2	2	1	2	2	2	2	2	2	2	2	2	23/24	95.8
Physical fitness assessment
Floria et al. (2018) ⁴⁶	2	2	2	2	2	2	2	0	2	2	1	2	21/24	87.5
López-Cuervo et al. (2025) ⁴⁷	2	2	2	2	2	2	2	0	-	-	-	-	14/16	87.5
Andrade et al., (2012) ⁴⁸	2	1	1	2	2	2	2	0	-	-	-	-	12/16	75.0
Laffaye et al., (2014) ⁴⁹	2	1	1	2	2	2	2	0	-	-	-	-	12/16	75.0
Gál-Pottyondy et al., (2025) ⁵⁰	2	2	1	2	2	2	2	0	2	2	2	2	21/24	87.5
Gómez- Carmona et al., (2021) ⁵¹	2	2	1	2	2	2	2	0	2	2	2	2	21/24	87.5
Teramoto et al., (2018) ⁵²	2	1	1	2	2	2	2	0	-	-	-	-	12/16	75.0
Stojanovic et al., (2018) ⁵³	2	2	1	2	2	2	2	0	-	-	-	-	13/16	81.3
Toselli et al., (2025) ⁵⁴	2	2	1	2	2	2	2	0	2	2	2	2	21/24	87.5
Panoutsakopoulos et al. (2014) ⁵⁵	2	2	1	2	2	2	2	0	2	2	2	2	21/24	87.5
Keogh et al., (2023) ⁵⁶	2	2	2	2	2	2	2	0	-	-	-	-	14/16	87.5
Irid et al., (2025) ⁵⁷	2	2	2	2	2	2	2	0	2	2	2	2	22/24	91.7

Note. The MINORS checklist comprises 12 items: (1) a clearly stated aim, (2) inclusion of consecutive patients, (3) prospective collection of data, (4) endpoints appropriate to the aim of the study, (5) unbiased assessment of the study endpoint, (6) follow-up period appropriate to the aim of the study, (7) loss to follow-up less than 5%, (8) prospective calculation of the study size, (9) an adequate control group, (10) contemporary groups, (11) baseline equivalence of groups, (12) adequate statistical analyses. Items 1–8 apply to all studies; items 9–12 apply only to comparative studies.

Common strengths across all studies included clearly stated aims (item 1: 100% adequate reporting), appropriate endpoints (item 4: 100%), unbiased endpoint assessment (item 5: 100%), appropriate follow-up periods (item 6: 100%), and adequate statistical analyses (item 12: 100% of comparative studies). The most frequent limitation was the lack of prospective sample size calculation (item 8), which was adequately reported in only three studies (9.4%). Other notable deficiencies included inconsistent reporting of prospective data collection (item 3: 43.8% adequate) and consecutive patient inclusion (item 2: 65.6% adequate). Among comparative studies (n = 19), all demonstrated adequate control groups (item 9), contemporary groups (item 10), and adequate statistical analysis (item 12), with only one study showing partial deficiency in baseline equivalence reporting (item 11). It should be noted that MINORS assesses study design rigor but does not evaluate PCA-specific methodological practices, which are reported separately in Section 3.3.4.

Main outcomes

The results are organized into three performance domains, enabling a comparative interpretation of PCA application across different basketball contexts.

PCA in workload monitoring

PCA was primarily applied to reduce external load metrics into interpretable components reflecting basketball physical demands. The workload monitoring domain included ten studies with sample sizes ranging from 11 to 94 participants (Table 3). All studies included exclusively male basketball players, with ages spanning from 17.6 years in youth elite populations to 38.48 years in elite referees. Competitive levels varied from collegiate to elite professional leagues including the Spanish ACB, NCAA Division I, and international competitions. Study settings encompassed both training sessions and competitive matches, with monitoring periods ranging from single tournaments to full seasons. The primary aims focused on characterizing position-specific load profiles, integrating multiple load dimensions, examining technological applications for load monitoring, and identifying key workload variables during competition.

Table 3.

Data extraction of the studies realized in basketball related to external and internal load in training and competition.

Author (year) country	SampleSexAgeCompetitive levelSetting	Aim	PCA methodology							PCA results
			Factorability	Data Suitability Testing		Retention Criteria			Rotation method	N Initial variables	N factors	% Variance explained	N variables extracted
			Retention Loading Criteria	Sample Adequacy Criteria	Sphericity Criteria	Factor	Loading	Cross-Loading	Rotation method	N Initial variables	N factors	% Variance explained	N variables extracted
Salazar et al. (2020) ²⁶Country: Spain	N = 17 (6 G, 4 F, 7 C)Sex: MaleAge: 23–28 yearsLevel: Elite (Basketball Club Baskonia)Setting: 5 friendly games during pre-season.	To describe specific demands to generate position-related time motion profiles in elite basketball	NR	KMOG: 0.72F: 0.47C: 0.68	NR	Kaiser (eigenvalue > 1)	>0.7	NR	Varimax	8	3	72.44	PC1 (tACC, hACC, tCOD, hCOD); PC2 (hJUMP); PC3 (tDEC, hDEC)
											4	75.38	PC1 (tCOD, hCOD); PC2 (tACC); PC3 (hDEC); PC4 (hJUMP)
											3	66.93	PC1 (tACC, hACC); PC2 (tJUMP, hJUMP); PC3 (hDEC)
Saucier et al. (2021) ²⁷Country: United States	N = 15 (7 G, 4 F, 4 C)Sex: MaleAge: 21.20 ± 1.80 yearsLevel: NCAA Division ISetting: 23 weeks during 2019–2020 season (77 training, 35 games)	To present the use of Strive Sense3 smart compression shorts to measure external load (accelerometry) and muscle activation (sEMG) during a season of training and competition for an NCAA Division I men's basketball team	NR	KMO > 0.5 for all positions	Bartlett's test – all values significant (p < 0.05)	Kaiser (eigenvalue > 1)	>0.7	NR	Varimax	8	Game
											G: 2	75.58	PC1 (tDistance, jumps, tACC, tDEC, hACC, hDEC); PC2 (SPEEDavg, SPEEDmax)
											F: 2	87.71	PC1 (tACC, tDEC, hACC, hDEC, jumps); PC2 (tDIST, SPEEDavg, SPEEDmax)
											C: 2	77.66	PC1 (tDistance, jumps, tACC, tDEC, hACC, hDEC); PC2 (SPEEDavg, SPEEDmax)
											Training
											G: 2	70.42	PC1 (tDistance, jumps, tACC, tDEC, hACC, hDEC); PC2 (SPEEDavg, SPEEDmax)
											F: 2	71.08	PC1 (tDistance, tACC, tDEC, hACC, hDEC, jumps); PC2 (SPEEDavg, SPEEDmax)
											C: 2	75.52	PC1 (jumps, tACC, tDEC, hACC, hDEC); PC2 (tDistance, SPEEDavg, SPEEDmax)
Zhang et al. (2026) ²⁸Country: China	N = 16 (7 G, 4 F, 4 C)Sex: MaleAge: 21.20 ± 1.80 yearsLevel: CollegiateSetting: 20 official games	To investigate the relationships between well-being, recovery state, game loads, and game performance across official games	NR	KMO > 0.50	Bartlett's test – all values significant (p < 0.01)	Kaiser (eigenvalue > 1)	>0.6	NR	Varimax (orthogonal rotation)	26	6	79.00	PC1 – Loads (Total COD, left/right COD, explosive actions, PlayerLoad, ACC/DEC, IMA events, sRPE)
													PC2 – Well-being (fatigue, sleep, stress, mood)
													PC3 – Jump (Total jump, low jump, medium jump)
													PC4 – High-intensity jump
													PC5 – Recovery (TQR score)
													PC6 – PlayerLoad per minute
García-Rubio et al. (2024) ²⁹Country: Spain	N = 12 (2 PG, 2 SG, 4 SF, 2 PF, 2 C)Sex: MaleAge: NRLevel: Elite level (ACB league)Setting: 10 training sessions over two microcycles (only 5-on-5 full-court tasks)	To determine the load profiles of players according to internal and external load values	NR	KMO = 0.81	Bartlett's test (p < 0.001)	Kaiser (eigenvalue > 1)	>0.6	Highest Loading	Varimax	21 (19 external load and 2 internal load)	4	81.5	PC1 – Versatile player (tDIST, ExpDIST, DIST_6−12km/h, DIST_{12−18 km/h}, hACC, hDEC, tIMP, hIMP, PL)
													PC2 – Explosive player (ACCmax, DECmax, HRmax, HRavg, SPEEDavg, SPEEDmax)
													PC3 – Dynamic player (DIST_0−6km/h, tACC, tDEC)
													PC4 – Fast Player (DIST_18−21km/h, DIST_21−24km/h)
Ibañez et al. (2025) ³⁰Country: Europe (different teams)	N = 94 (34 G, 47 F, 13 C)Sex: MaleAge: 17.6 ± 0.8 yearsLevel: Elite U-18 (EuroLeague Basketball ANGT finals)Setting: 2017–2018 ANGT finals tournament	To identify the main variables that explain basketball players’ physical load according to quarters playing positions using principal component analysis.	r > 0.70	KMO = 0.64–0.78	Bartlett's test – all values significant (p < 0.05)	Kaiser (eigenvalue > 1)	>0.6	Highest Loading	Varimax (orthogonal rotation)	31	5	61.1	PC1 – Explosive profile (ExpDIST, tDEC_2−3m/s, tIMP_3−5G)
													PC2 – Speed changes profile (tACC, tACC_0−1m/s, tDEC_0−1m/s)
													PC3 – Vertical jump profile (tLAND_3−5G, tLAND_5−8G, tTAKEOFF_3−5G, tTAKEOFF_5−8G)
													PC4 – Interior player profile (tIMP_3−5G, rightCENTFmax, leftCENTFmax)
Svilar et al. (2018) ³¹Country: Spain	N = 13 (4 G, 4 F, 5 C)Sex: MaleAge: 25.0–26.3 ± 2.2–4.1 yearsLevel: Elite professional level (ACB league)Setting: 2016–2017 season, 16 weeks (2 training sessions per week)	To investigate the structure of interrelationships among external and internal training session loads and determine how they vary among different positions using principal component analysis	NR	KMOG: 0.84F: 0.85C: 0.85	Bartlett's test – all positions significant (p < 0.01)	Kaiser (eigenvalue > 1)	>0.7	NR	Varimax (orthogonal rotation)	8	G: 3	100	PC1 (tACC, hACC, tCOD, hJUMP); PC2 (tDEC, tJUMP); PC3 (hDEC, hCOD)
											F: 3	100	PC1 (tACC, tDEC, hDEC, tCOD, hCOD); PC2 (hACC, tJUMP); PC3 (hJUMP)
											C: 2	100	PC1 (tACC, tDEC, tCOD, hCOD, hJUMP); PC2 (hACC, hDEC, tJUMP)
Rojas-Valverde et al. (2020) ³²Country: Europe (different countries)	N = 12Sex: MaleAge: 38.48 ± 3.47 yearsLevel: Elite level FIBA basketball refereesSetting: Euroleague Basketball ANGT finals 2017	To ascertain referees’ main external workload variables and analyze the effect of match periods (quarters) and accumulated points difference by quarter during a European youth basketball championship	r > 0.70	KMO > 0.50	Bartlett's test – all values significant (p < 0.05)	Kaiser (eigenvalue > 1)	>0.6	Highest Loading	Varimax (orthogonal rotation)	44	Q1: 5	80.56	PC1 (HSR Abs, DIST_18−21km/h, DIST_>24km/h, ACCmax, SPEEDmax)
											Q2: 4	79.71	PC1 (DIST_0−6km/h, SPEEDmax, tIMP_3−5G)
											Q3: 4	74.82	PC1 (ExpDIST, tACC, ACCmax, tACC_1−2m/s, tACC_2−3m/s, tACC_4−5m/s)
											Q4: 3	74.69	PC1 (DIST_>24km/h, tACC, tACC_1−2m/s)
Stone et al. (2022) ³³Country: United States	N = 11 (4 G, 3 F, 4 C)Sex: MaleAge: NRLevel: NCAA Division-I Power-5 conferenceSetting: 27 games during 2020–2021 NCAA season (24 regular phase, 1 conference tournament, 2 NCAA tournament)	To utilize principal component analysis (PCA) to simplify external load metrics from IMU data collected by playing positions	NR	KMO = 0.81	Bartlett's test – all values significant (p < 0.01)	Kaiser (eigenvalue > 1)	>0.7	NR	Varimax (orthogonal rotation)	109	2	81.30	PC1 – Volume/Intensity profile (tDEC, SPEEDavg, tACC, tMECH, tJUMP)
Stone et al. (2022) ³³Country: United States			NR	KMO = 0.81	Bartlett's test – all values significant (p < 0.01)	Kaiser (eigenvalue > 1)	>0.7	NR	Varimax (orthogonal rotation)	109	2	81.30	PC2 – Maximum speed profile (SPEEDmax)
Zhang et al. (2025) ³⁴Country: China	N = 16 (4 G, 3 F, 4 C)Sex: MaleAge: 27.2 ± 1.4 yearsLevel: Chinese National Basketball LeagueSetting: 2024 season, 18 games	To examine the contribution of external-internal load, perceived recovery status, and well-being to PIR and to determine how these factors influence player performance at various individualized performance levels	NR	KMO > 0.50	Bartlett's test – all values significant (p < 0.01)	Kaiser (eigenvalue > 1)	>0.6	NR	Varimax (orthogonal rotation)	26	6	83.00	PC1 – Composite load (sRPE, PL, mACC, hACC, mDEC, hDEC, mCODleft, hCODleft, mCODright, hCODright, mExpACTIONS, hExpACTIONS, mIMA, hIMA)
													PC2 – Well-being (total well-being score (fatigue, sleep, muscle soreness, stress, mood)
													PC3 – Composite Jump Load (tJUMP, lJUMP, mJUMP)
													PC4 – PlayerLoad minute (PL/min)
													PC5 – Muscle Soreness (individual index)
													PC6 – Recovery Status (TQR score)
Pino-Ortega et al. (2022) ³⁵Country: Europe (different teams)	N = 94 (34 G, 47 F, 13 C)Sex: MaleAge: 17.6 ± 0.8 yearsLevel: Elite U-18 (EuroLeague Basketball ANGT finals)Setting: 2017–2018 ANGT finals tournament	To set kinematic behavior parameters during official matches by principal component analysis (PCA), to examine distribution by birth dates and to analyze the relative age effect by playing positions	r > 0.70	KMO > 0.50	Bartlett's test – all values significant (p < 0.05)	Kaiser (eigenvalue > 1)	>0.6	Highest Loading	Varimax (orthogonal rotation)	250	3	66.3	PC1 – Volume load (tLAND_8−100G, tDIST)
													PC2 – Speed Changes (ACCmax, ACCavg)
													PC3 – Jump performance (TAKEOFFavg, LANDavg)

Note: KMO, Kaiser-Meyer-Olkin measure of sampling adequacy; N, number of participants; NR, not reported; PC, principal component; PCA, principal component analysis; r, correlation coefficient; RAE, relative age effect; %, percentage. Player Positions: C, centers; F, forwards; G, guards; PF, power forwards; PG, point guards; SF, small forwards; SG, shooting guards. Competition Levels: ACB, Asociación de Clubes de Baloncesto (Spanish professional basketball league); FIBA, Fédération Internationale de Basketball; NCAA, National Collegiate Athletic Association. Performance Metrics: PIR, Performance Index Rating; sRPE, session rating of perceived exertion; TQR, Total Quality Recovery scale. Distance & Speed Variables: ExpDIST, explosive distance; HSR Abs, high-speed running absolute; RD, relative distance; SPEEDavg, average speed; SPEEDmax, maximum speed; tDIST, total distance. Acceleration & Deceleration Variables: ACCavg, average acceleration; ACCmax, maximum acceleration; DECmax, maximum deceleration; hACC, high-intensity accelerations; hDEC, high-intensity decelerations; mACC, medium-intensity accelerations; mDEC, medium-intensity decelerations; tACC, total accelerations; tDEC, total decelerations. Changes of Direction: COD, changes of direction; hCOD, high-intensity changes of direction; mCOD, medium-intensity changes of direction; mCODleft, medium-intensity changes of direction to the left; mCODright, medium-intensity changes of direction to the right; hCODleft, high-intensity changes of direction to the left; hCODright, high-intensity changes of direction to the right; tCOD, total changes of direction; maxCENTFleft, left maximum centripetal force; maxCENTFright, right maximum centripetal force. Jump Variables: hJUMP, high-intensity jumps; lJUMP, low-intensity jumps; mJUMP, medium-intensity jumps; LANDavg, average landing duration; tJUMP, total jumps; tLAND, total landings; TAKEOFFavg, average take-off duration; tTAKEOFF, total take-offs. Impact & Load Variables: hExpACTIONS, high-intensity explosive actions; hIMP, high-intensity impacts; hIMA, high-intensity inertial movement analysis (IMA) events; mExpACTIONS, medium-intensity explosive actions; mIMA, medium-intensity IMA events; PL, PlayerLoad; PL/min, PlayerLoad per minute; tIMP, total impacts; tMECH, total mechanical load. Internal load: HRavg, average heart rate; HRmax, maximum heart rate; sEMG, surface electromyography.

Regarding PCA outcomes, the number of initial variables ranged from 8 to 250, with extracted component solutions varying from 2 to 6 factors that explained 61.1% to 100% of total variance. Component composition ranged from single-variable components to comprehensive multi-variable structures containing up to 14 variables. Despite this methodological variability, consistent structural patterns emerged. The first principal component (PC1) predominantly captured composite load or volume-intensity profiles, integrating total and high-intensity accelerations and decelerations, total distance, explosive distance, and PlayerLoad metrics. Additionally, PC1 frequently included impacts, changes of direction, and jump-related variables. PC2 most represented speed-related variables (average and maximum speed), though alternative structures included well-being indicators (fatigue, sleep, stress, mood), jump-specific metrics, or explosive performance characteristics (maximum acceleration, deceleration, and heart rate).

PCA in key performance indicators

PCA was predominantly used to identify latent structures underlying official game statistics. The key performance indicators domain comprised ten studies with sample sizes ranging from 32 national teams to 8511 player-season observations (Table 4). Most studies focused on male players, with two studies examining female athletes and one including both sexes. All studies analyzed elite-level competition, including NBA, WNBA, Spanish ACB League, EuroLeague Basketball, and FIBA international tournaments. Study settings predominantly featured official league or tournament matches, with monitoring periods spanning from single tournaments to multiple seasons (up to 23 seasons). The primary aims focused on developing player evaluation models, examining performance structure in competitive contexts, identifying factors influencing selection processes, establishing normative performance indicators, and creating unified frameworks for roster construction and team performance assessment.

Table 4.

Data extraction of the studies realized in basketball related performance indicators during competition.

Author (year) country	SampleAgeCompetitive levelSetting	Aim	PCA methodology							PCA results
			Factorability	Data Suitability Testing		Retention Criteria			Rotation method	N Initial variables	N factors	% Variance explained	N variables extracted
			Retention Loading Criteria	Sample Adequacy Criteria	Sphericity Criteria	Factor	Loading	Cross-Loading	Rotation method	N Initial variables	N factors	% Variance explained	N variables extracted
Oliveira & Newell (2024) ³⁶Country: United States	N = 900 playersSex: MaleAge: 28.73 ± 4.46 yearsLevel: Elite (National Basketball Association)Setting: 4920 regular season games across 4 seasons (2015–2016 to 2018–2019)	Develop ON score for evaluating athlete performance using hierarchical structure and PCA	NR	NR	NR	All retain	All retain	NR	Unrotated	22	22	100	Highest positive weights: AST, 2PTm, FTm, 3PTm, PTS
Oliveira & Newell (2024) ³⁶Country: United States			NR	NR	NR	All retain	All retain	NR	Unrotated	22	22	100	Lowest negative weights: BLKa, BLK, Age, Height.
Ke et al. (2024) ³⁷Country: United States	N = 8511 player-season observationsSex: Male and FemaleLevel: Elite (NBA and WNBA)Setting: 23 regular seasons (30 teams, 2000–2022)	Develop a unified machine learning framework for basketball team roster construction using both unsupervised and supervised learning methods	Absolute value of coefficients used to identify significant features	NR	NR	Total EVR > 0.9	> 0.9	Top 3 features	Unrotated	71	13	91%	PC1, Offensive ability (MIN; 2PTm; 2PTa)
													PC2, Traditional “big player” (OREB; DREB; REB)
													PC3, Westbrook-style playmaker (3PTm; AST%; ASTr)
													PC4, Secondary PG role (NRTG; AST; USG)
													PC5, Second-best defender (MIN; +/-; DRTG)
													PC6, poisoning 3-pointer tendency (2PT%, DREB, e2PT%)
da Silva & Rodrigues (2022) ³⁸Country: United States	N = 3006 player-season observationsSex: MaleAge: NRLevel: Elite (NBA)Setting: NBA regular seasons from 2013–14 to 2018–19 (6 seasons)	Identify factors considered by voters in All-NBA Teams selection and suggest a new selection format using PCA to evaluate players’ performances	NR	NR	NR	Total EVR > 0.9	NR	NR	Unrotated	15	2	74.9%	PC1, Performance (PTS, MIN, FT, FTa, 2PT, 2PTa, TOV, DREB)
da Silva & Rodrigues (2022) ³⁸Country: United States			NR	NR	NR	Total EVR > 0.9	NR	NR	Unrotated	15	2	74.9%	PC2, Position-related playing style
Doroshenko et al. (2020) ³⁹Country: Europe (different national teams)	N = 96 playersSex: FemaleAge: NRLevel: Elite national teams (EuroBasket 2017, final top-8 teams)Setting: 52 official matches from EuroBasket Women 2017 (Spain, France, Belgium, Greece, Turkey, Latvia, Italy and Slovakia)	To examine, analyze, and generalize the competitive activity structure based on the hierarchy of technical-tactical indicators in won and lost games using PCA	r > 0.5	NR	NR	All retain	> 0.5	NR	Varimax rotation	16	Won: 4	Won: 67.4%	PC1, Ball Control (REB, DREB; OREB; BLK)
													PC2, Long-Dist Offensive Actions (3PTm, 3PT% STL, AST, MIN)
													PC3, Game Offensive Actions (FTm, FT%; 2PTm, 2PT%, PTS)
													PC4, Turnovers (FLr, TOV)
											Lost: 4	Lost: 69.52%	PC1, Long-Distance Points (3PTm, 3PT%, PTS, MIN)
													PC2, Ball Control (REB, OREB, DREB, BLK)
													PC3, Efficient Assists (STL, AST)
													PC4, Game Offensive Actions (FTm, FT%, 2PTm, 2PT%)
Sampaio, et al., (2010) ⁴⁰Country: Spain	N = 198 players (5309 records)Sex: MaleAge: NRLevel: Elite professional – ACB League (Spain)Setting: 306 matches from the 2007–2008 regular season	To identify within-season differences in basketball players’ game-related statistics according to team quality and playing time	r > 0.4	KMO = 0.78	Anti-image correlation matrix revealed all variables above acceptable level of 0.5	Kaiser (eigenvalue > 1)	> 0.4	NR	Varimax (orthogonal rotation)	11	5	82%	PC1, Free-Throws (FTm, FTnm)
													PC2, 2PT FG (2PTm, 2PTnm)
													PC3, 3PT FG (3PTm, 3PTnm)
													PC4, Passing (AST, FLr)
													PC5, Errors (TOV, FLc)
Contreras et al., (2025) ⁴¹Country: Europe (different teams)	N = 1,919,273 actions, 25,032 timeoutsSex: MaleAge: NRLevel: Elite (EuroLeague Basketball)Setting: EuroLeague basketball games across 15 seasons (2008–2009 to 2022–2023)	To evaluate hierarchical clustering methodologies (agglomerative & divisive) for identifying patterns in timeout requests, using PCA for dimensionality reduction	NR	NR	NR	Kaiser (eigenvalue > 1)	3 PC retained	NR	Unrotated	21	3	69.64%	PC1, Scoring & Duration (MIN, PTS, 2PTm, 2PT, 3PTm, 3PT, FG)
													PC2, Efficiency – Ball management (AST, OREB, DREB, TOV)
													PC3 – Free Throws & Fouls (FTm, FTnm, FLc)
Péndola- Reinecke et al., (2025) ⁴²Country: Europe (different teams)	N = 386 playersSex: FemaleAge: NRLevel: Elite (Women's EuroLeague)Setting: Official FIBA match statistics from 410 games across 3 seasons (2021–2024)	To explore new dimensions in the classification of positions in women's basketball through a comprehensive statistical approach using PCA and k-means clustering.	r > 0.3	KMO = 0.722	Bartlett's test – all values significant (p < 0.01)	Kaiser (eigenvalue > 0.9)	> 0.30	NR	Unrotated	21 (per minute)	7	82.1%	PC1, General Offensive Production (FG, FGm, PTS, 2PT, 2PTm, FT, FTm)
													PC2, Rebounding & Interior Defence (REB, DREB, OREB, BLK)
													PC3, Perimeter Shooting (3PT, 3PTm, 3PT%)
													PC4, Field-Goal Efficiency (2PT%, FG%)
													PC5, Defensive Disruption & Assist (STL, TOV, AST)
													PC6, Fouls (FLc)
													PC7, Free-throw percentage (FT%)
Katris (2023) ⁴³Country:World (different national teams)	N = 32 national teamsSex: MaleAge: NRLevel: Elite (FIBA World Cup)Setting: FIBA World Cup 2019	To quantitatively investigate team performance using PCA, assess shoots, rebounds, turnover, & FT contributions compare offense vs. defense, and propose MLM to improve Power Rankings predictions.	NR	NR	NR	EVR > 0.8	NR	NR	Unrotated	3	1	80%	PC1 (Win percentage, Point difference, Final ranking)
Katris (2023) ⁴³Country:World (different national teams)			NR	NR	NR	EVR > 0.8	NR	NR	Unrotated	3	1	80%	PCA was used to create a single “Team Score” metric that represents overall team performance by combining three dimensions: win percentage, point difference, and final ranking.
Yin (2014) ⁴⁴Country: United States	N = 80 players (10 players from 8 teams)Sex: MaleAge: NRLevel: Elite (NBA)Setting: NBA season 2011–2012	To develop a player evaluation model using PCA and factor analysis for NBA players, calculating comprehensive ability scores and their relationship with salary via regression analysis.	NR	KMO = 0.79	Bartlett's test – all values significant (p < 0.01)	Kaiser (eigenvalue > 1)	> 0.7	Highest loading	Varimax (orthogonal rotation)	10	3	89.48%	PC1, Attack factors (AST, FLc, PTS, STL, MIN, Games played)
													PC2, Defence factors (REB, BLKa)
													PC3, Stable factors (FG%, FT%)
Courel-Ibañez et al., (2025) ⁴⁵Country: Spain	N = 609 playersSex: FemaleAge: NRLevel: Elite (LF Endesa, Spanish First Division)Setting: 10 seasons (2012–2022), 1834 official league games	To establish normative performance indicators for professional women's basketball using data-driven player role classification, identifying performance clusters by level (High, Mid, Low) to inform coaching, recruitment, and team management.	r > 0.5	KMO = 0.82	Bartlett's test – all positions significant (p < 0.01)	Kaiser (eigenvalue > 1)	> 0.5	NR	Varimax (orthogonal rotation)	15 (per minute)	4	75%	PCA + clusters analysis
													PC1, Secondary post (2PTm, OREB, FLa)
													PC2, Primary post (2PTm, OREB, DREB, BLKm, AST, FLa)
													PC3, Role player (Balanced profile)
													PC4, Playmaker (AST, STL, 3PTm)
													PC5, Versatile (Top in most stats, and FLc, FLa, 3PTm, AST)
													PC6, Specialist (3PTm, 3PTnm, FLa, BLKa)

Note. N = sample size; PCA = Principal Component Analysis; EVR = Explained Variance Ratio; PC = Principal Component; NR = Not Reported; KMO = Kaiser-Meyer-Olkin measure; NBA = National Basketball Association; WNBA = Women's National Basketball Association; ACB = Spanish Basketball League; MIN = minutes played; PTS = points; AST = assists; REB = rebounds; OREB = offensive rebounds; DREB = defensive rebounds; FG = field goals; FGm = field goals missed; 2PT = 2-point field goals; 2PTm = 2-point field goals made; 2PTa = 2-point field goal attempts; 3PT = 3-point field goals; 3PTm = 3-point field goals made; 3PTa = 3-point field goal attempts; 3PTnm = 3-point field goals not made; FT = free throws; FTm = free throws made; FTa = free throw attempts; FTnm = free throws not made; TOV = turnovers; STL = steals; BLK = blocks; BLKa = blocks against; BLKm = blocks made; FLc = fouls committed; FLr = fouls received; FLa = fouls against; USG = usage rate; +/- = plus/minus rating; NRTG = net rating; DRTG = defensive rating; AST% = assist percentage; ASTr = assist ratio; e2PT% = effective 2-point percentage.

Regarding PCA outcomes, studies analyzed between 3 and 71 performance indicators, though most focused on 10 to 22 variables. From these datasets, researchers extracted between 1 and 22 components accounting for 67.4% to 100% of the variance. Regarding component structure, the PC1 predominantly captured general offensive production or scoring ability, integrating points, field goals made (2-point and total), free throws made, and minutes played. The PC2 most represented rebounding and interior defense capabilities (total rebounds, defensive rebounds, offensive rebounds, blocks), though alternative structures included position-related playing style, efficiency in ball management (assists, turnovers), or traditional “big player” characteristics. Beyond these primary dimensions, additional components typically isolated specific dimensions such as perimeter shooting (3-point field goals), passing ability (assists), defensive disruption (steals), fouls, or shooting efficiency percentages.

PCA in physical fitness assessment

PCA served to identify the dimensional structure of neuromuscular and conditioning variables across basketball populations. The physical fitness assessment domain comprised twelve studies with sample sizes ranging from 12 to 273 participants (Table 5). Most studies examined male athletes, with four studies focusing on female populations and one including both sexes. Age ranged from 11.4 years in youth academy players to 31 years in professional athletes, with most studies concentrating on adolescent and young adult populations. Competitive levels varied from amateur to elite professional, including NBA Draft Combine participants, national team athletes, university varsity teams, and youth academy players from professional clubs. Study settings predominantly featured laboratory-based assessments or controlled testing sessions, including countermovement jump tests on force plates, on-court physical fitness batteries, change of direction speed protocols, and longitudinal monitoring across competitive seasons. The primary aims focused on identifying biomechanical determinants of vertical jump performance, examining relationships between neuromuscular variables and game performance, developing physical fitness profiles, establishing reliability and validity of assessment protocols, and predicting future performance based on fitness characteristics.

Table 5.

Data extraction of the studies realized in basketball related to physical conditioning.

Author (year)Country	SampleAgeCompetitive levelSetting	Aim	PCA methodology							PCA results
			Factorability	Data Suitability Testing		Retention Criteria			Rotation method	N Initial variables	N factors	% Variance explained	N variables extracted
			Retention Loading Criteria	Sample Adequacy Criteria	Sphericity Criteria	Factor	Loading	Cross-Loading	Rotation method	N Initial variables	N factors	% Variance explained	N variables extracted
Floria et al. (2018) ⁴⁶Country: Spain	N = 34 (17 IG, 17 CG)Sex: FemaleAge: 18–28 yearsLevel: AmateurSetting: Pre-post training intervention (6 weeks), countermovement jump tests	Determine effects of training intervention on force-, velocity-, and displacement-time curves using PCA to identify improvements in vertical jump performance	NR	NR	NR	95% explain var	NR	NR	Unrotated	F-T curve: 8	6	96	Pre-post changes: PC4 (Force at down-up transition, force at end of up movement); PC5 (force at end of up movement)
										V-T curve: 8	4	96	Pre-post changes: PC2 (Peak down velocity during CMJ); PC3 (Peak take-off velocity)
										D-T curve: 8	2	97	Pre-post changes: PC2 (rate of descent during CMJ and CMJ depth – peaks during and at end of down movement)
López-Cuervo et al. (2025) ⁴⁷Country: United States	N = 12Sex: MaleAge:18.3 ± 0.9 yearsLevel: Elite – U20 national-level professional athletesSetting: Two microcycles (deload: 3 days; overload: 5 days with 4 official matches)	Examine relationships between biomarkers, CMJ performance, and psychometric measures (session-RPE, wellness) across deload/ overload microcycles to inform allostatic load index development in team-sport athletes.	NR	KMO = 0.545–0.694	Bartlett'sχ² = 112,p < 0.001	NR	> 0.50	Highest loading	Varimax (orthogonal)	15	2	67.5%	PC1, Physiological stress (CK, EWMA session-RPE, Cumulative wellness)
López-Cuervo et al. (2025) ⁴⁷Country: United States			NR	KMO = 0.545–0.694	Bartlett'sχ² = 112,p < 0.001	NR	> 0.50	Highest loading	Varimax (orthogonal)	15	2	67.5%	PC2, Acute neuromuscular fatigue (CMJ post-workout, Session-RPE)
Andrade et al., (2012) ⁴⁸Country: Brazil	N = 19Sex: FemaleAge: 26.2 ± 4.7 yearsLevel: Elite – National Team (Olympic and FIBA Americas Championship)Setting: Training period for Copa América 2009	Investigate the contribution of biomechanical parameters (force, velocity, displacement-time curves) for vertical jump performance	NR	NR	Bartlett's test – all values significant (p < 0.01)	Kaiser (eigenvalue > 1)	NR	NR	Varimax (orthogonal)	CMJ: 8	CMJ: 2	CMJ: 73.39%	PC1 (PFPa, TPFPa, LR, Texc, TPFP)PC2 (PFP, TDF, Tcon)
Andrade et al., (2012) ⁴⁸Country: Brazil			NR	NR	Bartlett's test – all values significant (p < 0.01)	Kaiser (eigenvalue > 1)	NR	NR	Varimax (orthogonal)	CMJrun: 9	CMJrun: 3	CMJrun: 79.15%	PC1 (PFPa, TPFPa, LR, PFP)PC2 (Texc, TPFP, TDF)PC3 (Velmédia, Tcon)
Laffaye et al., (2014) ⁴⁹Country: United States	N = 273Sex: Male (n = 189) Female (n = 84)Age: 23–31 yearsLevel: Elite / College & Professional athletesSetting: Laboratory testing session; 6 maximal CMJs on force plate (500 Hz).	To assess the contribution of ECC-RFD, CON-F and time variables to CMJ performance, evaluate gender differences, and identify sport-specific PCA profiles	NR	NR	NR	Kaiser (eigenvalue > 1)	> 0.7	NR	Varimax (orthogonal)	5	2	76.8%	PC1, Time (Time, ECC-T)Athletes with longer contact and eccentric times = slow jumpersPC2, Force (R-ECC-RFD, CON-F, ECC-T:T)Athletes with high ECC-RFD and CON-F and low ECC-T:T = explosive jumpers
Gál-Pottyondy et al., (2025) ⁵⁰Country: Hungary	N = 23; (PCA sample = 14).Sex: FemaleAge: 15.22 ± 0.82 yearsLevel: Youth elite (Hungarian U16 First League)Setting: Postural stability tests (plank + 1-min dynamic balance) + monitoring of an official match using WIMU PRO™ system	To examine the relationship between postural stability and game-related performance, and to create PCA-derived competency scores	NR	NR	NR	Kaiser (eigenvalue > 1)	NR	Highest loading	Unrotated	10 (divided in 3 groups)	3	Anaerob. Endurance: 58.0%Explosiveactions: 51.3%Agility: 70.3%	PC1, Anaerobic Endurance: Expl. Distance/min; Speed AvgPC2, Explosive actionsAccMAX; DecMAXPC3, AgilityHIAccN/min; HIAccD/min; HIDecN/min; HIDecD/min
Gómez- Carmona et al., (2021) ⁵¹Country: Spain	N = 26Sex: 13 M, 13 FAge: M = 19.48 ± 1.41;F = 18.49 ± 2.27 yearsLevel: Semiprofessional (reserve teams of male & female 1st Spanish division teams)Setting: On-court physical fitness assessment across two testing sessions; battery including 6.75-m arc, unilateral jumps, Abalakov, RSA, 30–15 IFT and small-sided games	To identify sex-based physical fitness profiles through PCA and analyze the relationship with in-game roles	r > 0.60	KMOM = 0.657F = 0.623	Bartlett's test – all sex significant (p < 0.01)	Kaiser (eigenvalue > 1)	> 0.6	Highest loading	Varimax (orthogonal)	11	M: 4	M: 85.7%	PC1, Aerobic capacity & in-game physical conditioning (RSA Dec, 30–15 IFT, SSG PL_RT, SSG Dist, SSG Dist >16 km/h)
													PC2 – Bilateral jump & repeated jump capacity (Abalakov, Multijump, RSA, Acc)
													PC3, Curvilinear displacement (CentF left, CentF right)
													PC4 – Unilateral jump performance (right and left leg)
											F: 4	F: 83.6%	PC1 – Aerobic capacity & in-game physical conditioning (RSA Acc, 30–15 IFT, SSG PL_RT, SSG Dist, SSG Dist >16 km/h)
													PC2 – Jump capacity & curvilinear displacement (Abalakov, Multijump, CentF left, CentF right)
													PC3 – Unilateral jump performance (right and left leg)
													PC4 – Decelerative capacity (RSA Dec)
Teramoto et al., (2018) ⁵²Country: United States	N = 234Sex: MaleAge: 18–28 yearsLevel: Elite – NBA Draft Combine participantsSetting: NBA Draft Combine assessments (2010–2015) and subsequent NBA regular season performance (first year and 3-year averages)	To examine whether NBA Combine tests predict future NBA performance using PCA and principal component regression	NR	KMO = 0.789	Bartlett's test – all values significant (p < 0.01)	Kaiser (eigenvalue > 1)	> 0.5	Highest loading	Direct oblimin rotation (oblique)	13	3	72.4%	PC1, Length–Size (Wingspan, Hand length, Height, Standing reach, Hand width, Weight)
													PC2, Power–Quickness (Standing vertical jump, Maximum vertical jump, Three-quarter court sprint, Body fat %, Lane agility)
													PC3, Upper-Body Strength (Bench press)
Stojanovic et al., (2018) ⁵³Country: Serbia, Australia, Slovenia	N = 53 (29 backcourt, G, 24 frontcourt/ FW and C)Sex: MaleAge: 17.3 ± 1.0 (15–18 years)Level: Elite (Serbian Junior Basketball League)Setting: 5-day testing; 6 CODS tests; indoor hardwood court; electronic timing gates	To determine reliability, usefulness & factorial validity of 6 CODS tests, and evaluate positional differences	All CODS tests significantly correlated (r = 0.48–0.86)	NR	NR	Kaiser (eigenvalue > 1)	> 0.8	NR	Unrotated	6	1	74%	PC1, CODs Performance (Lane Arrow Closeout, Lane Agility Drill, Reactive Shuttle, Run-Shuffle-Run, Compass Drill, Modified 505)
Toselli et al., (2025) ⁵⁴Country: Italy	N = 153Sex: MaleAge: 11.4–17.8 years (U12-U19)Level: Elite youth academy (Serie A1)Setting: Anthropometry + BIVA in sport science laboratory (Jan 2022–May 2024)	To characterize morphological traits of elite young players using somatotype & BIVA; analyze age & maturation effects; develop PCA-based synthetic index	NR	NR	NR	Kaiser (eigenvalue > 1)	> 0.5	NR	Unrotated	3	1	80.74%	PC1, Somatotype composite (Endo, 0.648; Meso, 0.513, Ecto, −0.493)
Toselli et al., (2025) ⁵⁴Country: Italy			NR	NR	NR	Kaiser (eigenvalue > 1)	> 0.5	NR	Unrotated	3	1	80.74%	Somatotype composite = −1.66204 + (0.648 × endomorphy) + (0.5127 × mesomorphy) − (0.493 × ectomorphy)
Panoutsakopoulos et al. (2014) ⁵⁵Country: Greece	N = 173Sex: FemaleAge: 20.1 ± 2.8 yearsLevel: national-level athleteSetting: Laboratory-based SQJ testing on AMTI force plate (500 Hz); 3 maximal barefoot squat jumps without countermovement	To examine whether young adult female athletes from different sports rely on force- or time-dependent patterns during squat jump performance using PCA	r > 0.30	NR	NR	Kaiser (eigenvalue > 1)	> 0.50	NR	Varimax rotation	6	2	69.1%	PC1, Time (SBCM, tC, tFZmax)
Panoutsakopoulos et al. (2014) ⁵⁵Country: Greece			r > 0.30	NR	NR	Kaiser (eigenvalue > 1)	> 0.50	NR	Varimax rotation	6	2	69.1%	PC2, Force (FZbm, Pbm, RFDmax)
Keogh et al., (2023) ⁵⁶Country: Canada	N = 13Sex: FemaleAge: 20 ± 1.5 yearsLevel: University varsity basketballSetting: 17-week longitudinal CMJ monitoring (4 off-season, 7 pre-season, 6 in-season) using dual force plates (1000 Hz)	To compare week-to-week vs 7-week preseason reliability of CMJ metrics and PCA-derived components	Very high intercorrelation: each variable significantly correlated with 12–17 others r > 0.80	NR	NR	Kaiser (eigenvalue > 1)	NR	NR	Varimax rotation	18	6	92%	PC1, Force (Left Peak Propulsive Force, Left Peak Braking Force, Right Peak Propulsive Force)
													PC2, Braking (RSI mod, Average Braking RFD Asymmetry, Right Average Braking RFD)
													PC3, Power + JH (Jump Height, Peak Propulsive Power, Left Peak Landing Force)
													PC4, Asymmetry (Peak Landing Force Asymmetry, Peak Braking Force Asymmetry, Peak Propulsive Force Asymmetry)
													PC5, CMD + right landing (Right Peak Landing Force, CMD, Peak Braking Power)
													PC6, Strategy (RSI mod, CMD, Peak Landing Force Asymmetry)
Irid et al., (2025) ⁵⁷Country: France	N = 131Sex: MaleAge: 14.5 ± 0.7 yearsLevel: National youth pathway (FFBB)Setting: Comprehensive physical testing (speed, agility, jump, MAS) + retrospective tracking of adult performance levels.	To determine whether athletic profiles and relative age influence future achievement levels in young male basketball players	Strong intercorrelations; PCA used to reduce dimensionality for clustering	NR	NR	First two PCA axes	NR	NR	Unrotated	5	2	69.7%	PC1 (- association: Sprint, Butterfly; + association MAS, Jump height)
													PC2 (- association: MAS; + association: Stature & Jump)
													Clustering Results
													Cluster 1 “Hybrid": High MAS and jump height, lower sprint/Agility
													Cluster 2 “Elevated": Taller with good jumping, lower MAS
													Cluster 3 “Resilient": Shorter with lower jump but good MAS
													Cluster 4 “Explosive": Strong sprint/agility, lower jump/MAS

Note. CMJ, countermovement jump; CMJrun, countermovement jump preceded by run; Velmédia, average approach velocity; PFPa, passive peak force; TPFPa, time to passive peak force; LR, load rate; Texc, eccentric phase time; PFP, propulsion peak force; TPFP, time to propulsion peak force; TDF, rate of force development; Tcon, concentric phase time; PCA, principal component analysis; PC, principal component; KMO, Kaiser-Meyer-Olkin measure; λ, factor loading; R-ECC-RFD, relative eccentric rate of force development (RFD); CON-F, concentric force; TIME, total time; ECC-T, eccentric time; ECC-T:T, ratio of eccentric to total time; FZ_bm, Peak vGRF relative to body mass; P_bm, Peak power relative to body mass; RFD_max, Maximum rate of force development; t_C, Impulse time; t_FZmax, Time to achieve peak force; S_BCM, Vertical BCM trajectory during propulsion phase; CK, creatine kinase; EWMA, exponentially weighted moving average; RPE, rating of perceived exertion; F-T, force-time; V-T, velocity-time; D-T, displacement-time; IG, intervention group; CG, control group; RSA, repeated sprint ability; Dec, deceleration; Acc, acceleration; IFT, intermittent fitness test; SSG, small-sided games; PLRT, player load rate; Dist, distance; CentF, centripetal force; NBA, National Basketball Association; CODS, change of direction speed; BIVA, bioelectrical impedance vector analysis; Endo, endomorphy; Meso, mesomorphy; Ecto, ectomorphy; SQJ, squat jump; SBCM, vertical body center of mass trajectory during propulsion phase; JH, jump height; RSI mod, reactive strength index modified; CMD, countermovement depth; MAS, maximal aerobic speed; FFBB, Fédération Française de Basketball (French Basketball Federation); NR, not reported; M, male; F, female; G, guard; FW, forward; C, center.

Physical fitness studies included 3 to 18 variables, typically between 5 and 15 performance measures, yielding 1 to 6 components that accounted for 51.3% to 97% of variance. Component interpretation revealed a clear dimensional separation: the first component predominantly reflected temporal characteristics in vertical jump performance (contact time, eccentric time, time at peak force), aerobic capacity and in-game conditioning (repeated sprint ability, intermittent fitness test performance, distance covered), or morphological traits (somatotype, anthropometric dimensions). In contrast, the second component most often captured force or power attributes (peak force, rate of force development, maximum vertical jump, propulsive power), bilateral jump capacity, or braking-related metrics. Where additional components emerged, they isolated more specific qualities such as curvilinear displacement and change of direction ability, unilateral jump performance, inter-limb asymmetries, upper-body strength, or movement strategy patterns.

Methodological aspects of PCA

Regarding factorability assessment and data suitability testing, most studies (n = 22, 68.8%) did not report retention loading criteria such as correlation thresholds, with only ten studies (31.3%) establishing minimum correlation values ranging from r > 0.30 to r > 0.80. Without adequate correlation assessment, PCA may extract components reflecting random noise rather than meaningful underlying constructs. Sample adequacy was evaluated in thirteen studies (40.6%) using the Kaiser-Meyer-Olkin (KMO) measure, with values ranging from 0.47 to 0.85, predominantly exceeding the acceptable threshold of 0.50. KMO values below 0.50 are inadequate for PCA and values ≥0.60 are recommended, though most reporting studies demonstrated adequate sampling adequacy. Bartlett's test of sphericity was reported in fourteen studies (43.8%), with all achieving statistical significance (p < 0.01 or p < 0.05), confirming the suitability of correlation matrices for factor extraction.

Component retention criteria varied substantially, with the Kaiser criterion (eigenvalue > 1) being the most frequently applied method (n = 22, 68.8%), followed by cumulative variance explained thresholds ranging from 80% to 95% (n = 5, 15.6%), while five studies (15.6%) retained all components or used alternative criteria without explicit eigenvalue thresholds. Factor loading thresholds for variable inclusion ranged from > 0.30 to > 0.90, with > 0.50, > 0.60, and > 0.70 being the most common cut-off values applied in twenty-one studies (65.6%). Cross-loading management strategies were reported in only eight studies (25.0%) through “highest loading” criteria.

Regarding rotation, thirty-one studies (96.9%) reported their approach. Varimax orthogonal rotation was predominant (n = 20, 62.5%), followed by unrotated solutions (n = 11, 34.4%), with only one study employing oblique rotation. Varimax rotation is commonly applied in PCA as it enhances interpretability while maintaining the inherent orthogonality of extracted components. Unrotated solutions preserve the original mathematical structure but often compromise interpretability, whereas oblique rotation allows components to correlate when performance dimensions are theoretically interdependent, though this approach was rarely adopted.

Discussion

This systematic review sought to identify and synthesize studies applying Principal Component Analysis to basketball performance data across three primary domains (key performance indicators, workload monitoring, and physical fitness assessment), examining both current methodological practices and the inconsistencies that may affect the validity and reproducibility of findings. Despite the proliferation of multivariate data collection technologies in basketball and the increasing application of PCA as a dimensionality reduction technique, this review provides the first systematic examination of methodological quality, analytical approaches, and findings of PCA implementations in this sport. A total of 32 studies met the inclusion criteria. The majority (87.5%) demonstrated excellent methodological quality according to the MINORS assessment instrument.

The principal findings revealed consistent component structures within each performance domain, despite considerable methodological heterogeneity in PCA implementation. Workload monitoring studies predominantly identified composite load profiles integrating acceleration-deceleration dynamics as the primary performance dimension,^26–35 key performance indicator studies consistently extracted offensive production and rebounding-defense as distinct factors,^36–45 and physical fitness assessment studies differentiated time-dependent aerobic qualities from force-dependent power characteristics.^46–57 These findings provide evidence-based frameworks for organizing multidimensional basketball performance data and highlight both the utility and methodological challenges of PCA applications in sport science research.

Application of PCA in workload monitoring

Across workload monitoring studies, the first principal component consistently captured composite load or volume-intensity profiles. Total and high-intensity accelerations and decelerations, total distance, explosive distance, and PlayerLoad metrics were the variables most frequently integrated into PC1.^26–35 Previous research in team sports has demonstrated that acceleration-deceleration dynamics represent the primary dimension of neuromuscular demand during intermittent high-intensity activities,^58,59 and the present findings are consistent with this evidence. The repeated capacity to accelerate and decelerate appears to be a fundamental physical requirement in basketball, directly linked to the sport's characteristic movement patterns of rapid directional changes, defensive actions, and transition movements.^1,2 The secondary dimension most comprised speed-related variables and well-being indicators. High-velocity running, while relevant, constitutes a distinct and less prevalent demand compared to acceleration-based actions in basketball.^{26–28,33,34}

This hierarchical structure differs notably from sports with larger spaces and duration such as soccer or rugby, where total distance and high-speed running typically dominate the primary component.^14,60 It reflects the sport-specific differences in physical demand profiles. Several studies also incorporated well-being indicators (e.g., fatigue, sleep, stress, and mood) as secondary components,^28,34 consistent with contemporary frameworks emphasizing athlete readiness monitoring beyond purely mechanical metrics.⁶¹ Most studies, however, remained focused on objective wearable-derived variables. Internal load markers were notably underrepresented in the literature. Future research should integrate subjective measures, contextual factors such as training versus competition context, score differential, and playing time, alongside longitudinal adaptation patterns, to develop load-response models more useful for individualized training prescription in basketball populations.

Application of PCA in key performance indicators

The first principal component was consistent across key performance indicator studies. General offensive production — points, field goals made, free throws made, and minutes played — emerged as the dominant dimension in most analyses.^36–45 Scoring efficiency has long been recognized as the primary driver of individual player value and team success in basketball.⁶² Scoring-related variables also tend to exhibit the highest variance and strongest intercorrelations within performance datasets,⁶³ which partly accounts for their systematic dominance in PC1. The second component generally captured rebounding and interior defense capabilities, including total rebounds, defensive rebounds, offensive rebounds, and blocks, supporting theoretical models distinguishing between perimeter-oriented and interior-oriented contributions to team performance.⁶⁴

Additional components typically isolated specific dimensions such as perimeter shooting (3-point field goals), passing ability (assists), and shooting efficiency percentages.^37,42,44 These results indicate that modern basketball performance encompasses multiple semi-independent skill dimensions that cannot be adequately captured by unidimensional composite metrics, as players contribute to team success through genuinely different pathways such as high-volume scorers, defensive specialists, facilitating playmakers, or versatile multi-dimensional contributors.⁶⁵ This multidimensional structure has direct implications for roster construction and talent evaluation, as teams must balance complementary skill profiles rather than simply recruiting the highest-rated players according to single composite metrics. Most studies retained between two and seven components, explaining for 70–100% of total variance, suggesting that basketball performance can be reasonably represented by a small number of latent constructs. Nevertheless, the specific structures varied considerably depending on variable selection, competitive level, and analytical objectives. This highlights the context-dependent nature of performance organization and the limited utility of seeking universal taxonomies across all basketball contexts.

Application of PCA in physical fitness assessment

The physical fitness assessment domain revealed that the first principal component predominantly captured time-related characteristics in vertical jump performance (contact time, eccentric time, time at peak force), aerobic capacity and in-game physical conditioning (repeated sprint ability, intermittent fitness test performance, distance covered), or composite morphological traits (somatotype, anthropometric dimensions).^46–57 The prominence of temporal variables in vertical jump analysis^46,48,49,55 aligns with biomechanical research demonstrating that movement time and rate of force development represent fundamental determinants of explosive performance. Faster athletes have been characterized by shorter ground contact times and more rapid force application.⁶⁶ The emergence of aerobic capacity as a primary component in several studies^47,51 reflects the intermittent high-intensity nature of basketball, where the capacity to recover between high-intensity efforts and maintain performance across multiple quarters represents a critical physiological characteristic.^1,2

The second principal component most often represented force or power characteristics (peak force, rate of force development, maximum vertical jump, propulsive power),^48,49,55,56 supporting the theoretical distinction between time-dependent and force-dependent qualities in neuromuscular performance assessment.⁶⁷ This dimensional structure has important implications for talent identification and physical development programming, as it suggests that different athletes may achieve similar functional outcomes (e.g., vertical jump height) through different underlying strategies—some relying on rapid movement execution with moderate force production, others generating high forces over longer time periods.⁶⁸ The identification of additional components representing curvilinear displacement and change of direction ability,⁵¹ unilateral jump performance,⁵¹ and inter-limb asymmetries⁵⁶ highlights the multifaceted nature of basketball-specific physical fitness and the limitations of relying on single isolated tests to characterize athletic capabilities. These findings align with recent calls for comprehensive fitness assessment batteries that capture multiple performance dimensions and recognize that physical preparation strategies should be individualized based on athletes’ specific strength and weakness profiles across these dimensions.⁶⁹

Methodological quality and PCA implementation

The methodological quality assessment revealed that while most included studies demonstrated excellent overall rigor (87.5% scoring >75% on MINORS), substantial heterogeneity existed in PCA-specific methodological practices. The finding that 68.8% of studies failed to report correlation thresholds for factorability assessment and only 40.6% conducted sample adequacy testing (KMO) represents a major methodological concern, as these preliminary analyses are essential for determining whether PCA is an appropriate analytical technique for a given dataset.⁷⁰ The Kaiser-Meyer-Olkin measure and correlation matrices are fundamental prerequisites for determining PCA appropriateness,^17,18 yet their omission in most studies raises concerns about the validity of extracted components. The inconsistent application and reporting of these fundamental diagnostic procedures limits the interpretability of findings and may result in the extraction and interpretation of statistically derived components that lack substantive meaning or theoretical coherence.

The predominance of Varimax orthogonal rotation (62.5% of studies) reflects conventional practice in exploratory PCA applications, where the assumption of uncorrelated components facilitates interpretation and aligns with the mathematical properties of principal components.⁷¹ Rotation is essential to enhance the interpretability of components, and its absence or lack of justification may lead to ambiguous or unstable factor structures.¹⁸ However, the use of orthogonal rotation may be theoretically suboptimal in basketball contexts where performance dimensions are expected to exhibit correlations (e.g., offensive production and rebounding capabilities are not necessarily independent, as taller players with interior positioning often contribute to both domains).⁷¹ Only one study employed oblique rotation methods that permit component correlations,⁵² suggesting that researchers may be prioritizing mathematical convenience and interpretive simplicity over theoretical appropriateness.

Furthermore, the variability observed in the number of retained components may reflect inconsistent application of extraction criteria, such as the Kaiser criterion, which has well-documented limitations.¹⁷ To address this, researchers should employ multiple convergent retention methods (e.g., scree plot inspection, parallel analysis, variance thresholds) and explicitly justify retention decisions based on theoretical coherence rather than relying solely on eigenvalue-based criteria. The variation in factor loading thresholds (ranging from >0.30 to >0.90) and minimal reporting of cross-loading management strategies further highlight the lack of standardization in PCA implementation. These methodological inconsistencies complicate cross-study comparisons and synthesis, as different analytical decisions can substantially influence component structures, the number of components retained, and the substantive interpretation of results.⁷² Collectively, these findings underscore the need for standardized reporting guidelines for PCA in sports performance research.

Limitations and future research directions

Several limitations of this systematic review should be acknowledged. First, the restriction to peer-reviewed journal articles may have introduced publication bias, as studies with null findings or unsuccessful PCA applications are less likely to be published, potentially creating an overly optimistic representation of PCA utility in basketball research. Second, the heterogeneity in PCA methodology, variable selection, and sample characteristics precluded quantitative meta-analysis, preventing the establishment of pooled effect sizes, confidence intervals, or definitive benchmarks for component structures and limiting findings to qualitative consensus rather than quantitative evidence. Third, the MINORS instrument, while validated for non-randomized studies, assesses study design and conduct but does not evaluate PCA-specific methodological practices (e.g., factorability assessment, retention criteria, rotation justification), allowing studies to achieve excellent general quality scores despite substantial PCA implementation deficiencies. Fourth, the categorization of studies into three domains represents a simplification of basketball performance, as some studies integrated variables across multiple domains or examined hybrid constructs that do not fit neatly into discrete categories. Fifth, the heterogeneity of reporting standards across studies limited the depth of methodological comparisons.

Regarding future research, methodological standardization remains the most pressing priority. Researchers should systematically report factorability assessment procedures (e.g., correlation matrices, Kaiser-Meyer-Olkin measures, and Bartlett's test of sphericity) and justify component retention using multiple convergent criteria rather than relying exclusively on eigenvalue thresholds.¹⁹ Rotation method selection should follow theoretical reasoning about component independence, with oblique rotations considered when correlated dimensions are anticipated. Loading thresholds should be explicitly defined and cross-loading management strategies clearly documented. Beyond these procedural improvements, future investigations should move toward confirmatory approaches that test hypothesized structures across independent samples, examine solution stability over time, and validate PCA-derived scores against criterion performance measures.¹⁹ Longitudinal designs tracking how component structures evolve across developmental stages, training interventions, or competitive seasons would considerably advance current understanding of basketball performance dynamics and inform more individualized approaches to athlete development.

Conclusions and practical applications

This systematic review examined 32 studies applying PCA to basketball performance across workload monitoring, key performance indicators, and physical fitness domains. Overall study design quality was excellent (87.5% scoring >75% on MINORS), yet substantial heterogeneity existed in PCA-specific practices: 68.8% failed to report factorability assessment, 40.6% conducted sample adequacy testing, and 62.5% used Varimax rotation with minimal oblique methods. Despite methodological inconsistencies, consistent component structures emerged across domains, though the lack of standardization limits comparability and reproducibility of findings.

The identified component structures provide evidence-based frameworks for organizing multidimensional basketball performance data across three domains: (a) workload monitoring should prioritize acceleration-deceleration dynamics and total distance, complemented by speed and well-being measures; (b) performance evaluation should distinguish offensive production from rebounding-defensive contributions to guide talent identification and roster construction; (c) physical fitness assessment should differentiate time-dependent from force-dependent qualities requiring distinct training strategies.

To address the methodological variation identified, we propose a phase ordered PCA reporting checklist (Table 6) as a practical guideline for future research. This checklist synthesizes established best practices^17,18 with gaps identified in the present review, providing minimum reporting standards following the PCA workflow to standardize applications in sport research.

Table 6.

Methodological reporting standards for principal component analysis in sport research.

Phase	Item	Minimum standard	Reporting requirements
Phase 1: Data Preparation	Sample Size & Variables	N : variables ratio ≥ 5:1	Report sample size (N) and number of variables analyzed
Phase 1: Data Preparation	Data Standardization	Correlation matrix or z-scores	Report whether analysis used correlation or covariance matrix
Phase 2: Factorability Assessment	Correlation Matrix	r > 0.30 between variables	Report % of correlations > 0.30 or provide correlation matrix
	Kaiser-Meyer-Olkin (KMO)	KMO ≥ 0.60	Report KMO value with interpretation of sampling adequacy
	Bartlett's Test of Sphericity	p < 0.05	Report test statistics (χ²), degrees of freedom (df), and p-value
Phase 3: Component Extraction	Extraction Method	PCA specified	State “Principal Component Analysis” to distinguish from factor analysis
Phase 3: Component Extraction	Initial Eigenvalues	Report all eigenvalues	Provide eigenvalues and percentage of variance for all components
	Scree Plot	Visual inspection	Provide scree plot showing eigenvalue trajectory across components
Phase 4: Component Retention	Retention Criteria	Multiple convergent methods	Report ≥2 methods used (Kaiser criterion, scree plot inspection, parallel analysis, or cumulative variance threshold)
	Variance Explained	Cumulative ≥ 60%	Report % variance per component and cumulative total
	Theoretical Justification	Sport-context rationale	Explain how retained components relate to meaningful performance constructs
Phase 5: Rotation	Rotation Method	Method explicitly stated	Specify rotation type: Varimax, Promax, Oblimin, or unrotated
Phase 5: Rotation	Rotation Rationale	Theoretical justification	Explain choice based on expected component independence (orthogonal) or correlation (oblique)
Phase 6: Interpretation	Loading Threshold	\|λ\| ≥ 0.50	Explicitly state minimum threshold applied
	Component matrix	Rotated solution	Provide rotated component matrix showing salient loadings for each component
	Cross-Loading Management	Decision rule specified	State how variables loading on multiple components were assigned or excluded
	Component Labels	Sport-specific interpretation	Assign meaningful labels that reflect underlying performance constructs
	Component Composition	Variables with loadings	List variables defining each component with loadings
Reproducibility	Software Specification	Software and version	Report statistical software, version, and packages used

Footnotes

ORCID iDs

Carlos D. Gómez-Carmona

Ethical considerations

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Authors’ contributions

D.M-T and C.D.G-C conceptualized the study. J.R. and C.D.G-C developed the methodology. J.R. and D.M-T managed software implementation. S.J.I and C.D.G-C conducted formal analysis. J.R. and D.M-T performed the investigation. J.R. and C.D.G-C curated data. D.M-T and C.D.G-C prepared the original draft. J.R. and S.J.I reviewed and edited the manuscript. C.D.G-C and D.M-T created visualizations. J.R. and S.J.I supervised the project. S.J.I acquired funding. All authors have read and approved the final version of the manuscript.

Funding

This research was supported in part by: (a) Regional Government of Extremadura Research Group Support Grant (GR24133) from the Department of Education, Science, and Vocational Training, with 85% co-funding from the European Union via European Regional Development Funds (FEDER) and management oversight by the Spanish Ministry of Finance, (b) the Spanish National Research Agency through the project “Scientific and Technological Support for Analyzing Basketball Training Workload Based on Gender, Player Level, and Competitive Period” (PID2019–106614GBI00; MCIN/AEI/10.13039/501100011033), and (c) the Spanish Government's Sports Superior Council via the International Basketball Research Network (IBRN 20-24 and IBRN 20-25).

Consejo Superior de Deportes, Consejería de Economía, Ciencia y Agenda Digital, Junta de Extremadura, Agencia Española de Investigación, (grant number IBRN 20-24, IBRN 20-25, GR24133, MCIN/AEI/10.13039/501100011033, PID2019–106614GBI00). The author David Mancha-Triguero participates in this research due to the studies conducted during an international research stay at the Polytechnic Institute of Castelo Branco, funded by his university (CEU Fernando III University, CEU Universities)

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability

Not applicable.

References

Petway

Freitas

Calleja-González

, et al. Training load and match-play demands in basketball based on competition level: a systematic review. PLOS ONE 2020; 15: e0229212.

Russell

McLean

Impellizzeri

, et al. Measuring physical demands in basketball: an explorative systematic review of practices. Sports Med 2021; 51: 81–112.

Javadpour

Blakeslee

Khazaeli

, et al. Optimizing the best play in basketball using deep learning. J Sports Anal 2022; 8: 1–7.

Nguyen

DTA

, et al. The application of machine learning and deep learning in sport: predicting NBA players’ performance and popularity. J Inf Telecommun 2022; 6: 217–235.

Pino-Ortega

Rico-González

. The Use of Applied Technology in Team Sport. New York, NY: Taylor & Francis, 2021.

Chen

Rein

Memmert

, et al. A critical evaluation of data utilization and analytical techniques in basketball performance analysis: a systematic review. Int J Sports Sci Coach 2025; 20: 2286–2303.

Mănescu

. Big Data Analytics Framework for Decision-Making in Sports Performance Optimization. Data 2025; 10. Epub ahead of print 14 July 2025. DOI: https://doi.org/10.3390/data10070116.

Exel

Dabnichki

. Precision Sports Science: What Is Next for Data Analytics for Athlete Performance and Well-Being Optimization? Appl Sci 2024; 14. Epub ahead of print 16 April 2024. DOI: 10.3390/app14083361.

Fox

Scanlan

Stanton

. A review of player monitoring approaches in basketball: current trends and future directions. J Strength Cond Res 2017; 31: 2021.

10.

Pino-Ortega

Rojas-Valverde

Gómez-Carmona

, et al. Training design, performance analysis and talent identification—A systematic review about the most relevant variables through the principal component analysis in soccer, basketball and rugby. Int J Environ Res Public Health 2021; 18: 2642.

11.

Weaving

Beggs

Dalton-Barron

, et al. Visualizing the complexity of the athlete-monitoring cycle through principal-component analysis. Int J Sports Physiol Perform 2019; 14: 1304–1310.

12.

Greenacre

Groenen

PJF

Hastie

, et al. Principal component analysis. Nat Rev Methods Primer 2022; 2: 100.

13.

Villarejo-García

Puche-Ortuño

Gómez-Carmona

, et al. Multivariate analysis of performance indicators in elite women’s futsal: A principal component approach to understanding game dynamics. Int J Sports Sci Coach 2025. Epub ahead of print 23 May 2025. DOI: https://doi.org/10.1177/17479541251343609.

14.

Parmar

James

Hearne

, et al. Using principal component analysis to develop performance indicators in professional rugby league. Int J Perform Anal Sport 2018; 18: 938–949.

15.

Sarlis

Tjortjis

. Sports Analytics: Data Mining to Uncover NBA Player Position, Age, and Injury Impact on Performance and Economics. Information 2024; 15. Epub ahead of print 21 April 2024. DOI: 10.3390/info15040242.

16.

O’Donoghue

. Principal components analysis in the selection of key performance indicators in sport. Int J Perform Anal Sport 2008; 8: 145–155.

17.

Jolliffe

. Principal Component Analysis. Springer-Verlag, 2002. Epub ahead of print 2002. DOI: 10.1007/b98835.

18.

Hair

Babin

Anderson

, et al. Multivariate Data Analysis. Cengage Learning, 2022.

19.

Rojas-Valverde

Pino-Ortega

Gómez-Carmona

, et al. A systematic review of methods and criteria standard proposal for the use of principal component analysis in team’s sports science. Int J Environ Res Public Health 2020; 17: 8712.

20.

Ibáñez

Feu

. Research designs in sports science. E-Balonmano Com J Sports Sci 2026; 22: 1–44.

21.

Page

McKenzie

Bossuyt

, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. Int J Surg 2021; 88: 105906.

22.

Rico-González

Pino-Ortega

Clemente

, et al. Guidelines for performing systematic reviews in sports science. Biol Sport 2022; 39: 463–471.

23.

Santos CM da

Pimenta CA de

Nobre

MRC

. The PICO strategy for the research question construction and evidence search. Rev Lat Am Enfermagem 2007; 15: 508–511.

24.

Tang

Zhang

, et al. Kappa coefficient: a popular measure of rater agreement. Shanghai Arch Psychiatry 2015; 27: 62–67.

25.

Slim

Nini

Forestier

, et al. Methodological index for non-randomized studies (MINORS): development and validation of a new instrument. ANZ J Surg 2003; 73: 712–716.

26.

Salazar

Castellano

Svilar

. Differences in external load variables between playing positions in elite basketball match-play. J Hum Kinet 2020; 75: 257.

27.

Saucier

Davarzani

Burch

, et al. External load and muscle activation monitoring of NCAA division I basketball team using smart compression shorts. Sensors 2021; 21: 5348.

28.

Zhang

, et al. Integrating subjective assessments and wearable-derived metrics: decoding in-game load-performance relationships in collegiate basketball players. J Sports Med Phys Fitness 2026; 66: 10–21.

29.

García-Rubio

López-Sierra

Reina

, et al. Internal and external load profiles of male basketball players during training. Proc Inst Mech Eng Part P J Sports Eng Technol 2024. Epub: Ahead of Print. Epub ahead of print 12 February 2024. DOI: 10.1177/17543371241231033.

30.

Ibáñez

Rico-González

Gómez-Carmona

, et al. Physical workload patterns in U-18 basketball using LPS and MEMS data: a principal component analysis by quarter and playing position. Sensors 2025; 25: 6253.

31.

Svilar

Castellano

Jukic

, et al. Positional differences in elite basketball: selecting appropriate training-load measures. Int J Sports Physiol Perform 2018; 13: 947–952.

32.

Rojas-Valverde

Gómez-Carmona

Oliva-Lozano

, et al. Quarter’s external workload demands of basketball referees during a European youth congested-fixture tournament. Int J Perform Anal Sport 2020; 20: 432–444.

33.

Stone

Merrigan

Ramadan

, et al. Simplifying external load data in NCAA division-I men’s basketball competitions: a principal component analysis. Front Sports Act Living 2022; 4: 795897.

34.

Zhang

Chen

, et al. Tracking what truly matters: the contribution of game load, recovery, and well-being to basketball performance. Int J Sports Physiol Perform 2025; 20: 806–815.

35.

Pino-Ortega

Gómez-Carmona

Nakamura

, et al. Setting kinematic parameters that explain youth basketball behavior: influence of relative age effect according to playing position. J Strength Cond Res 2022; 36: 820–826.

36.

de Paula Oliveira

Newell

. A hierarchical approach for evaluating athlete performance with an application in elite basketball. Sci Rep 2024; 14: 1717.

37.

Bian

Chandra

. A unified machine learning framework for basketball team roster construction: NBA and WNBA. Appl Soft Comput 2024; 153: 111298.

38.

Silva

Rodrigues

. All-NBA Teams’ selection based on unsupervised learning. Stats 2022; 5: 154–171.

39.

Doroshenko

Sushko

Shamardin

, et al. Analysis of the competitive activity structure of skilled female basketball players. Phys Educ Theory Methodol 2020; 20: 219–227.

40.

Sampaio

Drinkwater

Leite

. Effects of season period, team quality, and playing time on basketball players’ game-related statistics. Eur J Sport Sci 2010; 10: 141–149.

41.

Contreras

Portillo

Luna

JMF

. Evaluation of hierarchical clustering methodologies for identifying patterns in timeout requests in EuroLeague basketball. Math 2025; 13: 2414.

42.

Péndola-Reinecke

Jiménez-Sáiz

Cuesta

, et al. Exploring new dimensions in the classification of positions in women’s basketball: a statistical approach. Appl Sci 2025; 15: 6159.

43.

Katris

. Investigation of FIBA world cup 2019: evidence using advanced statistical analysis and quantitative tools. Eng Proc 2023; 39: 83.

44.

Yin

. Principal component factor analysis-based NBA player comprehensive ability evaluation research. J Chem Pharm Res 2014; 6: 2400–2405.

45.

Courel-Ibáñez

Piñar López

Contreras-García

, et al. Profiling high performers in elite women’s basketball: functional roles and normative benchmarks from 10 seasons of Spanish first division data. J Sports Sci 2025; 43: 2764–2775.

46.

Floría

Sánchez-Sixto

Harrison

. Application of the principal component waveform analysis to identify improvements in vertical jump performance. J Sports Sci 2019; 37: 370–377.

47.

López-Cuervo

Rojas-Jaramillo

García-Caro

, et al. Biochemical and perceptual markers of physiological stress during acute exercise overload in U20 elite basketball players. Stresses 2025; 5: 52.

48.

Andrade

Amadio

Serrão

, et al. Contribuição dos parâmetros biomecânicos para o desempenho de saltos verticais de jogadoras de basquetebol. Rev Bras Educ Física E Esporte 2012; 26: 181–192.

49.

Laffaye

Wagner

Tombleson

TIL

. Countermovement jump height: gender and sport-specific differences in the force-time variables. J Strength Cond Res 2014; 28: 1096–1105.

50.

Gál-Pottyondy

Pálya

Trzaskoma

, et al. Integrating trunk endurance, dynamic stability, and in-game performance analysis in youth elite basketball players. BMC Sports Sci Med Rehabil 2025; 17: 269.

51.

Gómez-Carmona

Mancha-Triguero

Pino-Ortega

, et al. Exploring physical fitness profile of male and female semiprofessional basketball players through principal component analysis—A case study. J Funct Morphol Kinesiol 2021; 6: 67.

52.

Teramoto

Cross

Rieger

, et al. Predictive validity of national basketball association draft combine on future performance. J Strength Cond Res 2018; 32: 396–408.

53.

Stojanović

Aksović

Stojiljković

, et al. Reliability, usefulness, and factorial validity of change-of-direction speed tests in adolescent basketball players. J Strength Cond Res 2019; 33: 3162–3173.

54.

Toselli

Zaccagni

Rinaldo

, et al. Somatotype and bioelectrical impedance vector analysis in the evaluation of reference characteristics of elite young basketball players. Appl Sci 2025; 15: 2894.

55.

Panoutsakopoulos

Papachatzis

Kollias

. Sport specificity background affects the principal component structure of vertical squat jump performance of young adult female athletes. J Sport Health Sci 2014; 3: 239–247.

56.

Keogh

JAJ

Bishop

Ruder

, et al. Tracking the reliability of force plate-derived countermovement jump metrics over time in female basketball athletes: a comparison of principal component analysis vs. conventional methods. J Sci Sport Exerc 2025; 7: 460–472.

57.

Irid

Pineau

J-C

De Larochelambert

, et al. Impact of athletic profiles and the relative age effect on the future achievement levels of young basketball players. Front Sports Act Living 2025; 7: 1616800.

58.

Delves

RIM

Aughey

Ball

, et al. The quantification of acceleration events in elite team sport: a systematic review. Sports Med - Open 2021; 7: 45.

59.

Harper

McBurnie

Santos

, et al. Biomechanical and neuromuscular performance requirements of horizontal deceleration: a review with implications for random intermittent multi-directional sports. Sports Med 2022; 52: 2321–2354.

60.

Oliva-Lozano

Rojas-Valverde

Gómez-Carmona

, et al. Impact of contextual variables on the representative external load profile of spanish professional soccer match-play: a full season study. Eur J Sport Sci 2021; 21: 497–506.

61.

Impellizzeri

Marcora

Coutts

. Internal and external training load: 15 years on. Int J Sports Physiol Perform 2019; 14: 270–273.

62.

Terner

Franks

. Modeling player and team performance in basketball. Annu Rev Stat Its Appl 2021; 8: 1–23.

63.

Cabarkapa

Deane

Fry

, et al. Game statistics that discriminate winning and losing at the NBA level of basketball competition. PloS One 2022; 17: e0273427.

64.

Ibáñez

Courel-Ibáñez

Contreras-García

, et al. Redefining player roles in professional women’s basketball: from traditional positions to functional profiles. PloS One 2025; 20: e0330726.

65.

Penner

LSJ

. Player archetypes within basketball: optimizing roster composition to create a championship team. Front Sports Act Living 2025; 7: 1639431.

66.

Weyand

Sternlight

Bellizzi

, et al. Faster top running speeds are achieved with greater ground forces not more rapid leg movements. J Appl Physiol 2000; 89: 1991–1999.

67.

Cormie

McBride

McCaulley

. Power-time, force-time, and velocity-time curve analysis of the countermovement jump: impact of training. J Strength Cond Res 2009; 23: 177–186.

68.

Samozino

Rejc

Prampero

, et al.

Optimal force-velocity profile in ballistic movements–altius: citius or fortius?

Med Sci Sports Exerc 2012; 44: 313–322.

69.

Morrison

Martin

Talpey

, et al. A systematic review on fitness testing in adult male basketball players: tests adopted, characteristics reported and recommendations for practice. Sports Med 2022; 52: 1491–1532.

70.

Field

. Discovering Statistics with IBM SPSS Statistics. Sage, 2013.

71.

Costello

Osborne

. Best practices in exploratory factor analysis: four recommendations for getting the most from your analysis. Pract Assess Res Eval 2005; 10. Epub ahead of print 1 January 2005. DOI: 10.7275/jyj1-4868.

72.

Fabrigar

Wegener

MacCallum

, et al. Evaluating the use of exploratory factor analysis in psychological research. Psychol Methods 1999; 4: 272–299.