ultras-global / README.md
cjerzak's picture
Update README.md
0da8268 verified
metadata
dataset:
  - name: ultras-groups-directory
tags:
  - sports
  - football
  - soccer
  - ultras
  - fan-culture
  - directory
license: bigscience-openrail-m

Ultras-Global Dataset

Hugging Face Datasets

Ultras Global: A Dataset of Football Fan Groups Amid Political and Societal Change, 1950-2025

This dataset, introduced in the paper by Bryan J. Simmons and Connor T. Jerzak, compiles comprehensive data on 106 ultras groups (organized hardcore football supporter collectives) across 36 countries on five continents. Ultras groups represent a dynamic intersection of sport, identity politics, and social mobilization, often emerging during periods of political transition, economic upheaval, or cultural shifts. The dataset draws from triangulated sources including Wikipedia, fan directories, news archives, and academic literature, spanning group formations from 1950 to 2025 (with verified founding years for 77 groups, median ~1990s).

Key features:

  • Temporal coverage: 1950–2025, with surges in the 1980s–2000s tied to democratization waves (e.g., post-Franco Spain, Arab Spring Egypt).
  • Geographic scope: Concentrated in Europe (e.g., Spain: 23 groups), Latin America (Chile: 7), and North Africa (Morocco: 5), with entries from Asia, Oceania, and beyond.
  • Variables: Founding/dissolution years, ideological leanings (e.g., left-wing, right-wing, apolitical), criminal reputation, bans/reasons, social media footprints (total followers: >10M across groups), rivalries, and club win rates (2018–2021).
  • Design: Extensible and auditable, with conservative coding (missingness treated as absence of evidence). Excludes club-owned associations (e.g., Brazil's torcidas organizadas) to focus on independent groups anchored in stadium sections.

The dataset enables analyses of ultras as amplifiers of societal tensions (e.g., nationalism, youth protest) or agents of unity, including quasi-experimental designs using close match outcomes to isolate effects of team success on mobilization.

Citation

If you use this dataset in your research, please cite the associated paper:

Simmons, Bryan J., and Connor T. Jerzak. 2025. "Ultras Global: A Dataset of Football Fan Groups Amid Political and Societal Change, 1950-2025." Preprint, University of Texas at Austin.

BibTeX:

@techreport{simmons_jerzak_ultras_2025,
  author = {Simmons, Bryan J. and Jerzak, Connor T.},
  title = {Ultras Global: A Dataset of Football Fan Groups Amid Political and Societal Change, 1950-2025},
  year = {2025},
  institution = {UT Austin},
  type = {Preprint}
}

For the full paper (preprint available upon request), contact the authors.

Dataset Structure

The dataset is provided as a single CSV file (ultras_global.csv) with 111 rows (106 core groups + 5 extensions). Columns include:

Column Description Example Value Notes
country Country of the club. "Spain" String; 36 unique values.
city Home city of the club. "Barcelona" String; missing for national teams.
club_id Unique numeric ID for the club. 29 Integer; 1–111.
club_name Official club name. "FC Barcelona" String.
club_name_alt Alternative/historical names or notes. "HSK Gradanski, HASK Gradanski" String; optional.
league Domestic league. "La Liga" String.
ultras_group Name of the primary ultras group. "Boixos Nois" String; core identifier.
stadium_section_capacity Approx. capacity of ultras' stadium section. "~3000" String; estimates like "~5000".
year_started Founding year (verified for 77 groups). 1981 Integer; 1950–2025.
year_ended Dissolution year (if applicable; n=15). 2003 Integer; missing if active.
ended_reason Reason for end (e.g., "banned", "internal_strife"). "Violence/New Barca President" String; qualitative.
banned_reason Reason for formal ban (n=12 groups banned). "gov_repression" String; e.g., "racism".
political_leaning Ideological orientation (n=55 classified; 29 left, 26 right). "Right Wing" String; e.g., "Left Wing", "Apolitical".
main_rival Primary rival club/group. "Espanyol Ultras" String; n=70.
criminal_reputation Qualitative reputation (low/moderate/high/extreme; n=80). "High" String.
agg_social_media_followers Total followers across platforms (median ~35K; total >10M). "235K" String; aggregated.
social_media_links URLs to social media (e.g., Instagram, Facebook). "https://www.instagram.com/boixosnois.fcbarcelona/" String; semicolon-separated.
inclusion Placeholder for diversity metrics (future expansion). "" String; currently empty.
win_rate_2018win_rate_2021 Club win % in season (subset with data; e.g., Chile groups). "41%" String; % format.

Summary Statistics:

  • Groups: 106+ (active + historical).
  • Countries: 36+ (top: Spain=23, Chile=7, Bosnia & Herzegovina=7, Croatia=6, Morocco=5, etc.).
  • Bans/Exits: 12 bans (e.g., violence, repression); 15 dissolutions.
  • Social Footprint: 10M+ total followers; median group ~35K.

For variable definitions and coding rules, see Appendix A in the paper (Table 1: Variable Definitions).

Usage

Load the dataset using the Hugging Face datasets library:

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("cjerzak/ultras-global")

# Access the data (single split: 'train' by default)
df = dataset['train'].to_pandas()

# Example: Filter Spanish ultras groups
spanish_groups = df[df['country'] == 'Spain']
print(spanish_groups[['ultras_group', 'political_leaning', 'year_started']])

Example Analysis: Temporal Trends

import pandas as pd
import matplotlib.pyplot as plt

# Load and clean founding years
df = dataset['train'].to_pandas()
df['year_started'] = pd.to_numeric(df['year_started'], errors='coerce')
founded = df.dropna(subset=['year_started'])

# Plot cumulative growth
founded.groupby('year_started').size().cumsum().plot(kind='line')
plt.title('Cumulative Ultras Group Formations (1950–2025)')
plt.xlabel('Year')
plt.ylabel('Cumulative Count')
plt.show()

For full reproducibility, the dataset schema supports merging with external sources (e.g., match data via club_name/league, macro indicators via country/year_started).

Sources and Methodology

  • Primary Sources: Wikipedia (group histories), fan sites (e.g., ultras directories), news (e.g., BBC, ESPN), academic works (e.g., Doidge 2020 on ultras diffusion).
  • Collection: Triangulated scraping + hand-coding; audit trails in free-text notes.
  • Operationalization: Independent, non-club-owned groups with stadium sections; excludes sponsored associations.
  • Updates: Living dataset; contributions welcome via pull requests (e.g., new groups, verified years).

See Section 3 of the paper for details.

Limitations

  • Coverage: Uneven (Europe/Latin America dominant; gaps in Asia/Africa); absence of evidence ≠ evidence of absence.
  • Missingness: Conservative (e.g., ideology for only 55 groups); noisy labels from sources.
  • Scope: Focuses on men's football; women's/diasporic extensions planned.
  • Causality: Descriptive + quasi-experimental (close matches); observational correlations limited.

We encourage extensions and corrections—contact us!

License

  • Dataset: CC-BY-4.0 (attribution required; see LICENSE).
  • Paper: Preprint; full rights reserved pending publication.

Contact

Questions? Open an issue on the Hugging Face repo. Contributions to expand coverage (e.g., Brazil, women's ultras) are especially welcome!