Exploring New Dimensions in the Classification of Positions in Women’s Basketball: A Statistical Approach

Abstract

The aim of this study was to explore new dimensions in the classification of positions in women’s basketball through a comprehensive statistical approach. A total of 386 players from the last three seasons (2021–2024) of the Women’s Euroleague were analyzed based on official performance statistics. Inclusion criteria required players to have participated in all three seasons, with a minimum of 20 min per game across at least five games per season. Using data from the last three seasons of the Women’s Euroleague, analysis of variance, principal component analysis, and k-means clustering were performed to identify specific playing patterns and roles. All performance indicators were normalized per minute to ensure comparability. ANOVA tests revealed significant statistical differences between traditional positions (p < 0.05), validating the relevance of positional analysis. PCA was then used to reduce dimensionality and extract the key performance components, while k-means clustering grouped players according to similar in-game behaviors. The results revealed significant differences between traditional positions (with a significance criterion of p < 0.05) and suggested the need for an updated position classification to better reflect the current dynamics of modern gameplay. According to Euroleague players’ performance, the cluster analysis revealed that three main roles emerged: “perimeter specialists”, “defensive specialists”, and “primary scorers and rebounders”. This reclassification highlights the increasing tactical and statistical complexity of women’s basketball, moving beyond rigid position labels. This new framework can positively influence training and competition strategies. It also provides coaches, analysts, and talent developers with a data-driven tool for roster optimization, role assignment, and game planning in elite-level women’s basketball.

1. Introduction

Over the past two decades, women’s basketball has undergone profound changes in technical demands, tactical sophistication, and physical conditioning [1,2,3]. This evolution is particularly evident in elite-level contexts such as the EuroLeague Women, where players no longer adhere strictly to traditional positional roles but instead assume diverse responsibilities based on situational game demands [4,5,6]. Unlike earlier models with rigid assignments, current strategies encourage player versatility, enabling frequent transitions between creation, execution, and defensive functions depending on team strategy and match flow [7]. These shifts are largely driven by changes in coaching philosophy, training methodologies, and the increasing competitiveness of international women’s leagues [8,9,10]. Recent literature highlights the emergence of “positionless basketball”, where performance is no longer assessed through static role definitions but rather through dynamic, context-dependent contributions on both ends of the court [11]. As tactical systems become more adaptive and player rotations become more fluid, traditional labels such as point guard, forward, or center may fail to reflect actual in-game behavior, especially in the women’s game. This creates a growing need for analytical tools capable of detecting functional roles based on objective game-related data rather than legacy nomenclature [12].

Traditionally, basketball has been structured around five positions: point guard (PG), shooting guard (SG), small forward (SF), power forward (PF), and center (C), commonly grouped into backcourt and frontcourt categories. These designations offer a foundational framework but fall short of capturing tactical variety, especially in leagues like the EuroLeague Women, where hybrid profiles are increasingly common [13,14]. The backcourt typically includes the point guard (PG) and shooting guard (SG), who are primarily responsible for ball handling, playmaking, and perimeter shooting. Conversely, the frontcourt comprises the small forward (SF), power forward (PF), and center (C), who operate closer to the basket and contribute to rebounding, interior defense, and post-scoring [15,16]. For instance, while the NBA emphasizes athleticism and individual creativity, producing hybrid players such as “stretch fours” or “point forwards” [17], European basketball has historically prioritized team-based execution, tactical rigor, and positional discipline [18]. Within women’s basketball, these contrasts become even more pronounced: the distribution of scoring, playmaking, and defensive duties tends to be more egalitarian, making the application of rigid position-based classifications less appropriate [19]. Moreover, contextual variables such as opponent quality, game location, and match dynamics further modulate performance outputs and role execution, complicating the interpretation of static labels [20,21,22]. This complexity underscores the analytical limitations of applying uniform positional schemes across genders and competitive contexts. Thus, a new perspective is needed: one that uses advanced analytics to define roles not by traditional nomenclature but by how players perform [11].

Despite the growing relevance of sports analytics, most performance-focused studies continue to prioritize men’s basketball, leading to the generalization of findings that do not account for sex-specific physiological, biomechanical, and tactical differences [23,24,25]. This is problematic, particularly as women’s basketball exhibits distinct structural and stylistic features that demand dedicated analysis [12,18]. For example, studies have shown that female players display different energy expenditure patterns, game rhythms, and tactical role distributions compared to their male counterparts [24]. Yet, performance classifications often remain grounded in male-derived frameworks, perpetuating outdated and potentially misleading interpretations. While some research has analyzed game-related statistics to differentiate winning and losing teams [26,27], few attempts have been made to use these metrics to redefine roles or classify players in a way that reflects the unique demands of women’s basketball [28]. This absence of a context-specific model hinders evidence-based decision-making for coaches, analysts, and team managers operating in high-performance environments.

To address these analytical shortcomings, multivariate statistical techniques such as principal component analysis (PCA) and unsupervised clustering (particularly k-means) offer robust alternatives. PCA reduces the complexity of datasets by transforming correlated performance variables into a smaller number of uncorrelated components, preserving most of the original variance while enhancing interpretability [29]. When applied properly, i.e., with appropriate normalization, component extraction criteria, and theoretical alignment, PCA can identify latent structures in player performance, paving the way for more accurate role classifications [29,30,31]. Once dimensionality is reduced, clustering algorithms such as k-means allow researchers to group players based on similarities in their statistical behavior, uncovering functional roles that transcend traditional position categories [11,32,33]. These unsupervised approaches have proven effective in identifying new tactical profiles in both male and female athletes, enabling the development of classification systems that reflect how players contribute across multiple dimensions of the game [22,28]. This combination of PCA and clustering forms the methodological core of the present study.

Thus, the aim of this research was to apply advanced statistical techniques to develop a more context-sensitive and functionally meaningful classification of player positions in women’s basketball. In particular, we sought to examine the variability of game-related performance indicators across traditional positions and to identify emergent roles through data-driven clustering. The specific objectives of this study were (i) to identify the most significant performance differences among the five classic basketball positions, (ii) propose a new classification system tailored to the EuroLeague Women, and (iii) define distinct role profiles within this new classification. We hypothesized that significant differences exist in the statistical profiles of players across traditional positions and that these differences when analyzed through PCA and clustering, would allow for a redefinition of roles better suited to the complexity of contemporary women’s basketball.

2. Materials and Methods

2.1. Study Design

This was a cross-sectional, quantitative, and retrospective study based on official performance statistics from the Women’s Euroleague. The primary goal was to examine positional differences and identify new performance-based roles using multivariate statistical techniques.

2.2. Participants and Inclusion Criteria

Player statistics for this study were sourced from the official FIBA website for the Women’s Euroleague [34].

Table 1. Participation data by season.

Season	Teams	Games	Players
2023–2024	16	136	230
2022–2023	21	135	266
2021–2022	20	139	253

Note. Distribution of teams, games, and players analyzed across the three seasons of the Women’s Euroleague.

summarizes the participation data by season.

Following the unification of data from all three seasons, the summary of participation (N = 410 games) is shown in

Table 2. Combined participation data.

Season	Teams	Games	Players
3	35	410	519

Note. Summary of participation data in the Women’s Euroleague during the three seasons studied, including the number of teams, games played, and players analyzed.

.

The dataset includes player statistics from the Women’s Euroleague over the last three seasons, obtained from official FIBA data. Players with meaningful participation across the three seasons were included.

This approach aimed to exclude samples whose participation in the competition was not significant, ensuring the reliability of the statistical data as a representation of the style of play in this tournament.

To enhance the analysis, we considered certain players who fulfilled dual roles, such as point guard/shooting guard or small forward/power forward. These players were included in the study to observe whether significant differences exist compared to the five traditional positions.

As shown in

Table 3. Summary of inclusion and participation criteria.

Position	Number of Players
Point Guard (PG)	78
Shooting Guarg (SG)	53
Guard (G)	34
Foward (F)	41
Small Foward (SF)	58
Power Foward (PF)	54
Center (C)	68
TOTAL	386

Note. Summary of selected players based on inclusion criteria (players must have participated in at least one of the three seasons under analysis, must have competed in at least 5 games, and must have an average of at least 20 min of playing time), showing the distribution of players by position and the dual roles reported by the teams.

, the final analysis includes a total of N = 386 players.

2.3. Ethical Considerations

The study was designed in compliance with the recommendations for clinical research outlined in the Declaration of Helsinki and approved by the Ethics Committee of the university (CIPI/19/095).

2.4. Variables

In

Table 4. Game-related statistics analyzed in the study.

Variable	Definition
P-T	Total points scored by a player
T2Pa	Two-point field goals made by a player
T2Pi	Two-point field goals attempted by a player
%T2P	Two-point field goal percentage of a player
T3Pa	Three-point field goals made by a player
T3Pi	Three-point field goals attempted by a player
%T3P	Three-point field goal percentage of a player
TCa	Total field goals made by a player
TCi	Total field goals attempted by a player
%TC	Field goal percentage of a player
TLa	Free throws made by a player
TLi	Free throws attempted by a player
%TL	Free throw percentage of a player
RO	Number of offensive rebounds grabbed by a player
RD	Number of defensive rebounds grabbed by a player
RT	Total number of rebounds grabbed by a player
AS	Number of assists made by a player
FP	Total number of fouls committed by a player
BP	Total number of turnovers a player has made against the defense
BR	Number of steals made by a player in defense
TP	Number of blocks made by a player against the opponent’s offensive shots

Note. Description of the game variables used in the statistical analysis of player performance in the Women’s Euroleague.

, all the variables analyzed in this study are presented, along with a definition for each. These variables are subsequently featured in other tables normalized per minute of play to ensure data consistency and proper analysis [35].

2.5. Data Sources

Official databases provided by the FIBA official website [34] were used, along with other recognized sports sources such as Basketmetrics [36], to collect additional data necessary for this study.

2.6. Data Validation

The data were validated through independent observation of 10 randomly selected games by two experienced analysts. Both analysts are graduates in Sports Sciences, certified FIBA Level 3 basketball coaches for over 5 years, and have more than 10 years of experience in the sport. Their metrics were compared with the data collected from the official FIBA database and Basketmetrics to ensure accuracy. Intra-class correlations were very good (ICC values > 0.81).

Specifically, a two-way mixed-effects model with absolute agreement was selected, as it accounts for both the specific rates used in the study and the need for exact matching of values.

The ICC values exceeded 0.81, indicating excellent reliability according to standard interpretation guidelines [37]. This supports the validity of the data used for subsequent statistical analyses.

2.7. Statistical Analysis

The statistical analysis in this study focuses on identifying and classifying new dimensions in high-level women’s basketball positions based on data from the last three seasons of the Women’s Euroleague. Several steps and statistical methods were employed to achieve this objective.

A descriptive analysis was conducted to examine the average performance of players based on their listed positions. While the traditional classification includes five positions (PG, SG, SF, PF, and C), teams often report hybrid roles (e.g., G or F). Therefore, seven positional categories were included in the analysis.

Firstly, variables were normalized per minute played. This step was crucial to ensure comparability among players and to minimize the effect of differences in playing time. Normalization adjusted all performance statistics proportionally, providing a more accurate basis for subsequent analysis [35].

Next, a descriptive analysis of traditional positions was conducted, offering an overview of the average characteristics and performance of the five classic basketball positions: point guard, shooting guard, small forward, power forward, and center [38]. This initial approach identified trends and variations within each position, providing contextual insights for more advanced analyses.

A one-way ANOVA with Games-Howell post hoc tests was then used to identify variables with significant differences (p < 0.05) among traditional positions. The Games-Howell tests were selected for their ability to handle variations in variances and unequal sample sizes, ensuring a robust evaluation of significant group differences.

Principal component analysis (PCA) was applied to reduce the dimensionality of the game-related statistics dataset. This technique transformed a set of potentially correlated variables into a set of uncorrelated variables, referred to as principal components. PCA identified the most significant variables contributing to differences in performance and player roles, providing a clearer understanding of the underlying dimensions of performance [27,39].

Finally, k-means clustering was performed to identify current player roles based on statistical analysis. This method grouped players into clusters according to similarities in their performance data, optimizing the number of clusters. The analysis revealed emerging patterns and roles among players beyond traditional positions, enabling a more updated and accurate classification [40].

Statistical analyses were conducted using Jamovi 2.5.3, an open-source statistical analysis platform. Jamovi was chosen for its user-friendly interface and robust capabilities for performing complex analyses efficiently. Results are shown as mean ± standard deviation, and the statistical significance threshold was set at p < 0.05.

3. Results

Table 5. Game-related statistic results.

POS	P/MIN	T2Pa/MIN	T2Pi/MIN	%T2P	T3Pa/MIN	T3Pi/MIN
C	0.349 ± 0.134	0.130 ± 0.0539	0.259 ± 0.0894	0.496 ± 0.0859	0.0118 ± 0.0152	0.0390 ± 0.0458
F	0.337 ± 0.125	0.0946 ± 0.0499	0.204 ± 0.0908	0.460 ± 0.127	0.0302 ± 0.0225	0.0963 ± 0.0560
G	0.315 ± 0.139	0.0782 ± 0.0364	0.174 ± 0.0676	0.454 ± 0.139	0.0336 ± 0.0255	0.116 ± 0.0548
PF	0.352 ± 0.131	0.103 ± 0.0544	0.212 ± 0.0889	0.469 ± 0.131	0.0294 ± 0.0229	0.0924 ± 0.0585
PG	0.275 ± 0.111	0.0660 ± 0.0371	0.159 ± 0.0696	0.407 ± 0.135	0.0319 ± 0.0174	0.108 ± 0.0443
SF	0.280 ± 0.122	0.0626 ± 0.0411	0.145 ± 0.0776	0.417 ± 0.118	0.0383 ± 0.0260	0.121 ± 0.0580
SG	0.292 ± 0.106	0.0623 ± 0.0323	0.142 ± 0.0687	0.433 ± 0.138	0.0421 ± 0.0266	0.140 ± 0.0548
POS	%T3P	TCa/MIN	TCi/MIN	%TC	TLa/MIN	TLi/MIN
C	0.242 ± 0.205	0.141 ± 0.0538	0.297 ± 0.0926	0.471 ± 0.0840	0.0565 ± 0.0336	0.0785 ± 0.0428
F	0.278 ± 0.145	0.125 ± 0.0479	0.301 ± 0.0805	0.410 ± 0.0954	0.0551 ± 0.0372	0.0754 ± 0.0509
G	0.251 ± 0.139	0.113 ± 0.0526	0.289 ± 0.0987	0.378 ± 0.0874	0.0555 ± 0.0364	0.0727 ± 0.0445
PF	0.283 ± 0.138	0.132 ± 0.0529	0.304 ± 0.0819	0.422 ± 0.103	0.0580 ± 0.0410	0.0770 ± 0.0517
PG	0.283 ± 0.101	0.0971 ± 0.0417	0.266 ± 0.0825	0.354 ± 0.0874	0.0488 ± 0.0355	0.0632 ± 0.0426
SF	0.294 ± 0.133	0.101 ± 0.0446	0.266 ± 0.0843	0.374 ± 0.0816	0.0412 ± 0.0302	0.0540 ± 0.0354
SG	0.275 ± 0.135	0.105 ± 0.0398	0.286 ± 0.0799	0.361 ± 0.0990	0.0423 ± 0.0278	0.0525 ± 0.0321
POS	TCa/MIN	TCi/MIN	%TC	TLa/MIN	TLi/MIN	%TL
C	0.141 ± 0.0538	0.297 ± 0.0926	0.471 ± 0.0840	0.0565 ± 0.0336	0.0785 ± 0.0428	0.702 ± 0.145
F	0.125 ± 0.0479	0.301 ± 0.0805	0.410 ± 0.0954	0.0551 ± 0.0372	0.0754 ± 0.0509	0.744 ± 0.113
G	0.113 ± 0.0526	0.289 ± 0.0987	0.378 ± 0.0874	0.0555 ± 0.0364	0.0727 ± 0.0445	0.761 ± 0.135
PF	0.132 ± 0.0529	0.304 ± 0.0819	0.422 ± 0.103	0.0580 ± 0.0410	0.0770 ± 0.0517	0.736 ± 0.157
PG	0.0971 ± 0.0417	0.266 ± 0.0825	0.354 ± 0.0874	0.0488 ± 0.0355	0.0632 ± 0.0426	0.735 ± 0.182
SF	0.101 ± 0.0446	0.266 ± 0.0843	0.374 ± 0.0816	0.0412 ± 0.0302	0.0540 ± 0.0354	0.753 ± 0.213
SG	0.105 ± 0.0398	0.286 ± 0.0799	0.361 ± 0.0990	0.0423 ± 0.0278	0.0525 ± 0.0321	0.794 ± 0.166
POS	RO/MIN	RD/MIN	RT/MIN	AS/MIN	FP/MIN	BP/MIN
C	0.0715 ± 0.0291	0.155 ± 0.0434	0.227 ± 0.0610	0.0521 ± 0.0246	0.105 ± 0.0378	0.0693 ± 0.0256
F	0.0522 ± 0.0355	0.132 ± 0.0599	0.185 ± 0.0829	0.0580 ± 0.0252	0.0971 ± 0.0434	0.0637 ± 0.0269
G	0.0279 ± 0.0246	0.0958 ± 0.0415	0.124 ± 0.0553	0.0888 ± 0.0498	0.0930 ± 0.0476	0.0755 ± 0.0243
PF	0.0578 ± 0.0300	0.143 ± 0.0474	0.201 ± 0.0628	0.0604 ± 0.0302	0.0917 ± 0.0407	0.0622 ± 0.0294
PG	0.0210 ± 0.0185	0.0840 ± 0.0339	0.106 ± 0.0427	0.122 ± 0.0516	0.0910 ± 0.0342	0.0814 ± 0.0324
SF	0.0312 ± 0.0184	0.101 ± 0.0467	0.133 ± 0.0522	0.0709 ± 0.0302	0.0938 ± 0.0392	0.0584 ± 0.0240
SG	0.0260 ± 0.0155	0.0783 ± 0.0363	0.104 ± 0.0375	0.0862 ± 0.0397	0.0960 ± 0.0510	0.0734 ± 0.0275
POS	BR/MIN	TP/MIN
C	0.0309 ± 0.0165	0.0251 ± 0.0174
F	0.0371 ± 0.0176	0.0149 ± 0.0169
G	0.0436 ± 0.0200	0.00788 ± 0.0105
PF	0.0337 ± 0.0202	0.0139 ± 0.0138
PG	0.0417 ± 0.0253	0.00372 ± 0.00723
SF	0.0390 ± 0.0177	0.00655 ± 0.00928
SG	0.0385 ± 0.0200	0.00396 ± 0.00531

Note. Description of the game variables used in the statistical analysis of player performance in the Women’s Euroleague. In columns labeled with percentage (%), the numerical value is expressed as a fraction (out of 1). C: Center. PF: power forward. SF: small forward. SG: shooting guard. PG: point guard. F: forward (power forward/small forward). G: guard (point guard/shooting guard).

displays the means and standard deviations of the normalized game variables (/MIN) for each playing position.

Table 6. One-way ANOVA results.

	F	gl1	gl2	p	η2
P/MIN	4.242	6	151	<0.001	0.027
T2Pa/MIN	17.607	6	152	<0.001	0.104
T2Pi/MIN	15.958	6	152	<0.001	0.095
%T2P	5.441	6	149	<0.001	0.035
T3Pa/MIN	17.378	6	147	<0.001	0.106
T3Pi/MIN	26.334	6	149	<0.001	0.153
%T3P	0.655	6	142	0.686	0.005
TCa/MIN	7.456	6	151	<0.001	0.047
TCi/MIN	2.135	6	151	0.052	0.014
%TC	14.500	6	151	<0.001	0.088
TLa/MIN	2.450	6	151	0.027	0.016
TLi/MIN	4.258	6	151	<0.001	0.027
%TL	1.763	6	144	0.111	0.012
RO/MIN	34.675	6	149	<0.001	0.189
RD/MIN	31.731	6	149	<0.001	0.176
RT/MIN	49.084	6	149	<0.001	0.248
AS/MIN	22.985	6	151	<0.001	0.132
FP/MIN	0.996	6	149	0.430	0.007
BP/MIN	5.012	6	153	<0.001	0.032
BR/MIN	2.930	6	152	0.010	0.019
TP/MIN	20.802	6	147	<0.001	0.124

Note. Results of the one-way analysis of variance (ANOVA) to identify significant differences among traditional positions, with a significance criterion of p < 0.05.

presents the results of the one-way ANOVA test, which was used to identify significant differences among traditional basketball positions in basic game variables. This statistical test evaluates whether there are differences in the means of these variables across positions.

The results revealed significant differences (p < 0.05) in most of the variables studied.

The results of the Games-Howell post hoc analysis revealed significant statistical differences among the traditional basketball positions in the Women’s Euroleague competition.

Significant differences between traditional playing positions were found in both two-point and three-point shooting metrics. Regarding two-point attempts per minute (T2Pa/MIN), centers attempted significantly more shots than all other positions (p < 0.001), with the largest differences observed compared to shooting guards and small forwards. Similarly, centers also registered the highest values in two-point shots made per minute (T2Pi/MIN), followed by power forwards. Guards (PG, SG, and G) had significantly lower values in both attempts and conversions.

On the other hand, players listed as shooting guards and small forwards showed significantly higher three-point attempt and conversion rates (T3Pa/MIN and T3Pi/MIN) compared to centers and power forwards (p < 0.001). These findings reflect expected role specializations, where perimeter players emphasize long-range shooting, and frontcourt players dominate in close-range scoring.

In terms of rebounding, significant differences were found across positions. Centers obtained the highest number of offensive rebounds per minute (RO/MIN), significantly surpassing all perimeter roles (p < 0.001). A similar trend was observed for defensive rebounds (RD/MIN), with centers and power forwards collecting more rebounds than guards. The total rebounding metric (RT/MIN) confirmed this pattern, with centers outperforming all other roles, particularly compared to point guards and shooting guards (p < 0.001). These results align with the physical positioning and tactical responsibilities assigned to frontcourt players.

Regarding assists per minute (AS/MIN), point guards registered significantly higher values than all other positions (p < 0.001), confirming their central role in game orchestration. The most notable differences were found between point guards and centers (mean difference = −0.07, p < 0.001) and between point guards and shooting guards (mean difference = −0.035, p < 0.001). Small forwards and power forwards recorded intermediate values, while centers had the lowest assist rates.

Table 7. Principal component analysis (PCA).

	Component							Unicity
	1	2	3	4	5	6	7
TCi/MIN	0.924		0.358					0.0930
P/MIN	0.847		0.303					0.0137
Tca/MIN	0.791			0.302				0.0293
T2Pi/MIN	0.790							0.0883
Tli/MIN	0.749						0.347	0.1542
T2Pa/MIN	0.730			0.307				0.0417
Tla/MIN	0.705						0.533	0.0821
RT/MIN		0.928						0.0776
RD/MIN		0.859						0.1946
RO/MIN		0.763						0.2739
TP/MIN		0.719						0.4370
T3Pa/MIN			0.965					0.0570
T3Pi/MIN			0.896					0.1323
%T3P			0.693					0.4265
%T2P				0.999				0.1676
%TC				0.855				0.0806
BR/MIN		0.344			0.883			0.3228
BP/MIN	0.346				0.656			0.3626
AS/MIN		−0.366			0.549	−0.425		0.2478
FP/MIN						0.956		0.1900
%TL							0.834	0.2881

Note. Results of the principal component analysis (PCA), highlighting the most significant variables.

presents a principal component analysis (PCA) conducted to reduce the dimensionality of the dataset and identify underlying structures in player performance. Prior to extraction, data suitability was verified using the Kaiser–Meyer–Olkin (KMO) test, which yielded a value of 0.722, and Bartlett’s test of sphericity, which was significant (p < 0.001), confirming the adequacy of the data for PCA.

Components with eigenvalues greater than 0.9 were retained, and the selection was further supported by the cumulative variance explained (above 82.1%). Factor loadings above 0.30 were used as the threshold for inclusion.

In this analysis, principal components are identified as linear combinations of the original variables, explaining a significant percentage of the total variance. The eigenvalues and percentages of variance explained indicate the relative importance of each component. For example, the first component explains 34.6% of the variance, and the second and third components explain 16.4% and 9.7%, respectively.

These principal components simplified the complexity of the data, enabling a better interpretation and understanding of player performance in women’s basketball, particularly in the Women’s Euroleague.

Figure 1 shows the clustering results in two dimensions. However, the optimal number of clusters (k = 3) was not determined from the visual representation but through statistical methods applied to the entire dataset. Specifically, the gap statistic was computed using the complete set of performance variables, not just the first two principal components. This approach ensures that the number of clusters reflects the full multidimensional structure of player performance. The use of a 2D cluster plot is intended solely for visual interpretation, where Dim 1 and Dim 2 serve as a projection of the original dataset, capturing over 50% of the total variance. Therefore, the three-cluster solution visualized in the figure is based on the full dataset and is not limited by the reduced dimensionality of the plot. These clusters represent distinct roles that players fulfill in women’s basketball beyond the traditional positions.

Each cluster groups players based on similarities in their performance statistics, enabling a new classification rooted in their specific contributions to the game.

Table 8. Player distribution across clusters.

Cluster No	Count
1	134
2	118
3	94

Note. Distribution of players across clusters based on the roles identified through k-means clustering analysis.

shows the distribution of players within each cluster proposed by the k-means clustering analysis. This distribution provides the basis for evaluating the significant differences among the groups identified.

To evaluate the statistical significance and practical magnitude of differences among the three clusters identified via k-means, a one-way Welch ANOVA was conducted for each performance variable. This was followed by Games-Howell post hoc comparisons.

Table 9. One-way clusters ANOVA results.

	F	gl1	gl2	p	η2
P/MIN	306.27	2	200	<0.001	0.754
T2Pa/MIN	172.89	2	200	<0.001	0.634
T2Pi/MIN	192.38	2	209	<0.001	0.648
%T2P	25.54	2	226	<0.001	0.184
T3Pa/MIN	169.61	2	216	<0.001	0.611
T3Pi/MIN	137.55	2	211	<0.001	0.566
%T3P	61.43	2	175	<0.001	0.412
TCa/MIN	273.67	2	200	<0.001	0.732
TCi/MIN	167.53	2	212	<0.001	0.612
%TC	79.54	2	211	<0.001	0.43
TLa/MIN	53.05	2	208	<0.001	0.338
TLi/MIN	68.17	2	206	<0.001	0.398
%TL	8.56	2	221	<0.001	0.072
RO/MIN	70.53	2	179	<0.001	0.441
RD/MIN	97.14	2	207	<0.001	0.484
RT/MIN	119.62	2	198	<0.001	0.547
AS/MIN	12.28	2	228	<0.001	0.097
FP/MIN	14.47	2	208	<0.001	0.122
BP/MIN	1.32	2	222	0.269	0.012
BR/MIN	1.12	2	221	0.329	0.01
TP/MIN	30.49	2	199	<0.001	0.235

Note. Results of the one-way analysis of variance (ANOVA) to identify significant differences among cluster positions, with a significance criterion of p < 0.05.

summarizes the F-values, degrees of freedom, and effect size calculations using eta squared (η²), providing a clearer understanding of how strongly each variable contributes to differentiating the clusters.

Figure 2 highlights the standout characteristics of players grouped in Cluster 1, referred to as “Perimeter Specialists”. These players excel in long-distance shooting and their ability to stretch opposing defenses, creating spaces for interior or paint players.

Figure 3 highlights the defining characteristics of players grouped in Cluster 2, referred to as “Defensive Specialists”.

Figure 4 highlights the defining characteristics of players grouped in Cluster 3, identified as “Primary Scorers and Rebounders”.

The Games-Howell post hoc comparisons revealed statistically significant differences (p < 0.001) across clusters in a wide range of performance variables. Specifically, Cluster 3 (“Paint Dominators and Primary Scorers”) consistently outperformed the other two in variables such as points per minute, two-point attempts, two-point makes, and total rebounds. These findings confirm the offensive and interior dominance of this group.

Conversely, Cluster 1 (“Perimeter Specialists”) showed significantly higher values in three-point attempts, three-point makes, and three-point percentage when compared to both Clusters 2 and 3, highlighting their specialization in long-range shooting. Additionally, Cluster 1 also contributed more to assists, reflecting their dual role as shooters and facilitators.

Cluster 2 (“Defensive Specialists”) was characterized by significantly higher values in personal fouls than Cluster 1, suggesting a more physically intense, defense-oriented style of play.

Some variables, such as turnovers and steals, did not show significant differences between clusters, indicating that they are not key discriminants of the roles identified.

These results support the functional and tactical distinctions captured by the cluster analysis and align with the practical demands of each identified role in elite-level women’s basketball.

4. Discussion

The primary objective of this study was to identify statistically significant differences across traditional playing positions in women’s basketball and to explore new role-based classifications using PCA and clustering. The hypothesis that performance indicators differ significantly between positions and that a new classification model can be developed was confirmed by the results.

The five traditional positions (point guard (PG), shooting guard (SG), small forward (SF), power forward (PF), and center (C)) showed considerable variations in their performance across most game statistics analyzed, showing the importance of this classification in the development of basketball performance. Anil Duman et al. [13] conducted a position-specific cluster analysis of basketball players and found that even within the five conventional roles, performance-based subgroups emerge, suggesting that the traditional classification may not fully capture the diversity of player behavior and contribution. Their findings support the use of data-driven models to reveal more accurate groupings based on actual game performance rather than fixed positional labels. Similarly, Richardson [31] analyzed the evolution of player roles in the NBA and highlighted how player classifications have become increasingly fluid over time. His work documents how hybrid roles, such as “point forwards” or “stretch bigs”, have emerged, reflecting tactical shifts and the influence of individual versatility. Together, these studies advocate for a dynamic and evidence-based re-evaluation of positional roles, which aligns directly with the objectives of the present research. PGs demonstrated outstanding results in assists per minute and generated a substantial number of points per minute, reflecting their essential role in playmaking and quick scoring in women’s basketball [4,32]. SGs showed high efficiency in three-point shooting and points per minute, highlighting their long-range scoring ability [4]. SFs exhibited a balance in defensive rebounds and assists, indicating their versatility in both defense and facilitating plays [4,13]. Power forwards (PFs) were above average in defensive rebounds and were closely matched with centers in blocks per minute, emphasizing their importance in defense and control of the paint [4,13,32]. Cs dominated in total rebounds and field goal efficiency, playing a crucial role in scoring near the basket and ball recovery [4,32].

The advanced statistical analysis conducted in this study unveiled new dimensions in the classification of positions in Women’s Euroleague basketball, emphasizing the importance of updating and expanding traditional methodologies for position classification [4,32]. This aligns closely with Baumann’s study, which re-examines men’s basketball positions using clustering techniques [33]. Like this research, Baumann utilized advanced data analysis techniques to redefine roles in men’s basketball, incorporating a greater number of variables and proposing continuous learning to understand the evolving dynamics of roles over time.

Cluster 1 identified “Perimeter Specialists”. These players had a crucial performance in long-distance shooting and their ability to stretch opposing defenses, creating spaces for interior or paint players. Their key attributes were as follows:

• High frequency and efficiency in three-point shooting: Demonstrated through metrics such as three-point attempts per minute (T3Pi/MIN) and three-point shooting percentage (%T3P).
• Significant contribution in assists: Highlighting their role in facilitating scoring opportunities for themselves and teammates.

This role is pivotal in breaking down defenses and creating scoring opportunities, both directly and indirectly. These players distinctly outperform others in metrics associated with perimeter play.

Cluster 2 identified “Defensive Specialists”. These players are important to the team’s defensive strategy and are tasked with neutralizing the primary offensive threats of opponents [4,32]. Their key attributes were as follows:

• Consistency in defensive rebounds: Highlighting their ability to secure possession and limit second-chance opportunities for the opposing team.
• Blocks and steals: Emphasizing their effectiveness in disrupting the opposing team’s offensive plays.
• High fouls per minute: These players showed the highest values in this metric compared to the other two clusters, reflecting their aggressive and impactful defensive style.

Defensive specialists are vital for maintaining a strong team defense, disrupting opponents’ rhythm, and safeguarding the team’s scoring potential by limiting the effectiveness of rival players.

Cluster 3 identified “Paint Dominators and Primary Scorers”. These players were the primary scoring contributors for their teams, with a crucial impact on rebounds [4,31,32]. Their key attributes were as follows:

• High points per minute (P/MIN): Demonstrating their consistent scoring ability and critical offensive presence.
• Total rebounds per minute (RT/MIN): Reflecting their effectiveness in securing both offensive and defensive rebounds, providing second-chance opportunities and limiting those of opponents.

These players combine reliable scoring with a strong rebounding presence, making them key contributors on both ends of the court. They stand out in metrics related to scoring efficiency and rebound dominance [31,32].

Overall, the results obtained in this study were largely in line with our expectations based on previous research and the practical understanding of playing roles in professional basketball. The emergence of Cluster 1 as “Perimeter Specialists” was anticipated, as players in backcourt positions, such as point guards and shooting guards, tend to exhibit higher values in three-point shooting and assists. Similarly, the identification of Cluster 3 as “Paint Dominators and Primary Scorers” aligns with the traditional dominance of frontcourt players, especially centers and power forwards, in close-range scoring and rebounding. The findings related to Cluster 2, the “Defensive Specialists”, were also consistent with expected roles, as these players showed higher values in personal fouls, suggesting an aggressive and disruptive defensive presence.

Nonetheless, some nuances were noteworthy. For example, the fact that players classified as guards contributed significantly to rebounding in certain clusters or that forwards exhibited higher assist rates than expected in traditional frameworks may reflect the increasing fluidity of roles in modern women’s basketball. These deviations from the classic positional expectations reinforce the need for updated, performance-based classifications that capture the evolving demands and versatility of elite female athletes.

Unlike Baumann’s study, this research focused on women’s basketball, an area with notably less existing literature [4,32]. As in Baumann’s work, principal component analysis (PCA) and k-means clustering were employed together to detect the linear relationships between game-related statistics (PCA) and to identify specific playing patterns and roles (clustering). These techniques allowed for the identification of nuanced player roles and patterns that are not fully captured by traditional classification systems.

In contrast, the study by Anıl Duman et al. [13] also used clustering methods but with the objective of identifying differences and defining roles within traditional position groups. Their findings highlighted that within each traditional position, players performed different roles and contributed to team performance through varied metrics. This broader approach provides a more comprehensive redefinition of roles within the court [4,13,31,32].

When comparing these findings to previous studies, it becomes evident that there is a growing trend toward updating traditional classifications through increasingly advanced statistical analyses to better define contemporary player roles [4,33].

Limitations and Practical Applications

The findings of this study have significant implications for both practice and theory in women’s basketball. From a practical standpoint, the identification of specific performance-based roles, such as perimeter specialists, defensive specialists, and primary scorers and rebounders, enables coaches and analysts to assign tasks more effectively, plan strategies tailored to each player’s strengths and optimize talent selection and development processes. These insights support the construction of rosters that maximize individual contributions and enhance overall team performance. Theoretically, this study contributes to the evolving framework of position classification by applying advanced statistical techniques, such as PCA and k-means clustering, to reflect modern gameplay more accurately. It moves beyond traditional static roles, offering a dynamic, data-driven model for evaluating player impact based on actual performance. Moreover, it addresses a critical gap in the literature by focusing specifically on women’s basketball, a context often overlooked in previous analyses. However, the main limitation of this study lies in its exclusive focus on the Women’s Euroleague, which, despite its high level of competition, may limit the generalizability of findings to other leagues or competitive contexts. Additionally, the absence of qualitative variables, such as tactical roles, coaching strategies, or game context, means that some dimensions of player behavior may not be fully captured by statistical outputs alone.

5. Conclusions

In addressing the primary objective, this study concludes that role classification in women’s basketball, using statistical techniques, effectively reveals the dynamics of modern gameplay. The results suggest that traditional positional classifications may no longer align with current styles of play and that a methodology based on principal component analysis (PCA) and k-means clustering provides a more precise and practical alternative.

In response to the specific objectives, the study finds that, despite significant differences observed among traditional positions, including variables such as assists, three-point shooting accuracy, and total rebounds, the three roles identified through clustering emphasize the importance of adopting more dynamic, data-driven approaches for classifying high-performance women’s basketball players. These roles were defined based on playing patterns:

• Perimeter specialists;
• Defensive specialists;
• Primary scorers and rebounders.

This role-based classification not only enhances the understanding of individual performance but also has the potential to positively impact training strategies, competition tactics, and decision-making processes in team roster formation.