Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Network analysis and link prediction in competitive women's basketball

Anthony Bonato, Morganna Hinds

TL;DR

This paper addresses predicting outcomes and future interactions in competitive women's basketball by constructing adversarial outcome networks for NCAA teams and node2vec embeddings for on-court interactions. It defines the CON Score $\mathrm{CON}(u)$ and the Low-Key Leader Score $\varepsilon_i$, and combines them with PageRank on reversed-edge graphs to identify influential but undercentral competitors. Node2vec embeddings are learned to capture higher-order structure and are evaluated for predicting NCAA March Madness matchups, future NCAA interactions, and WNBA blocking events, with pass predictions being weaker but interpretable. The results indicate network-derived features add predictive value beyond traditional statistics, and the work suggests integrating such features with covariates and extending to spatiotemporal data and other structurally similar team sports.

Abstract

Network structure and its role in prediction are examined in competitive basketball at the team and player levels. Adversarial game outcome networks from NCAA Division I women's basketball from 2021 to 2024 are used to compute the common out-neighbor score and PageRank, which are combined into a low-key leader strength that identifies competitors influential through structural similarity despite relatively low centrality. This measure is related to changes in NCAA NET rankings by grouping teams into quantiles and comparing average rank changes across seasons for both previous-to-current and current-to-next transitions. Link prediction is then studied using node2vec embeddings across three interaction settings. For NCAA regular-season game networks, cosine similarity between team embeddings is used in a logistic regression model to predict March Madness matchups. For WNBA shot-blocking networks, future directed blocking interactions are predicted via logistic regression on concatenated source-target player embeddings. For WNBA passing networks, region embeddings learned from first-quarter passes are evaluated for their ability to predict subsequent passing connections. Across NCAA and WNBA settings, embedding-based models provide statistically significant evidence that higher-order network structure contains predictive signals for future interactions, while the passing experiment shows weaker predictive performance but yields interpretable similarity patterns consistent with passing feasibility.

Network analysis and link prediction in competitive women's basketball

TL;DR

This paper addresses predicting outcomes and future interactions in competitive women's basketball by constructing adversarial outcome networks for NCAA teams and node2vec embeddings for on-court interactions. It defines the CON Score and the Low-Key Leader Score , and combines them with PageRank on reversed-edge graphs to identify influential but undercentral competitors. Node2vec embeddings are learned to capture higher-order structure and are evaluated for predicting NCAA March Madness matchups, future NCAA interactions, and WNBA blocking events, with pass predictions being weaker but interpretable. The results indicate network-derived features add predictive value beyond traditional statistics, and the work suggests integrating such features with covariates and extending to spatiotemporal data and other structurally similar team sports.

Abstract

Network structure and its role in prediction are examined in competitive basketball at the team and player levels. Adversarial game outcome networks from NCAA Division I women's basketball from 2021 to 2024 are used to compute the common out-neighbor score and PageRank, which are combined into a low-key leader strength that identifies competitors influential through structural similarity despite relatively low centrality. This measure is related to changes in NCAA NET rankings by grouping teams into quantiles and comparing average rank changes across seasons for both previous-to-current and current-to-next transitions. Link prediction is then studied using node2vec embeddings across three interaction settings. For NCAA regular-season game networks, cosine similarity between team embeddings is used in a logistic regression model to predict March Madness matchups. For WNBA shot-blocking networks, future directed blocking interactions are predicted via logistic regression on concatenated source-target player embeddings. For WNBA passing networks, region embeddings learned from first-quarter passes are evaluated for their ability to predict subsequent passing connections. Across NCAA and WNBA settings, embedding-based models provide statistically significant evidence that higher-order network structure contains predictive signals for future interactions, while the passing experiment shows weaker predictive performance but yields interpretable similarity patterns consistent with passing feasibility.
Paper Structure (11 sections, 5 equations, 9 figures, 2 tables)

This paper contains 11 sections, 5 equations, 9 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Centralities and Embeddings
  3. The CON Score
  4. The Low-Key Leader Score
  5. node2vec
  6. Experimental Design and Methods
  7. NCAA Women's Basketball Game Data
  8. WNBA Passing Data
  9. WNBA Blocking Data
  10. Results
  11. Discussion and Future Directions

Figures (9)

  • Figure 1: The structure of a March Madness regional bracket.
  • Figure 2: An NCAA 2023-2024 season network. Nodes with higher comparative out-degree are larger and darker.
  • Figure 3: Aces vs. Fever 09/21/25 passing network. Edges of higher weight are larger and darker in colour.
  • Figure 4: WNBA 2024 season blocking network. Nodes with higher comparative out-degree are larger and darker.
  • Figure 5: Average change in NET Ranking from 2023-2024, based on 2024 low-key leader strength.
  • ...and 4 more figures