theonlyduck/vglocations-public-preview

VGLocations Public Preview Dataset

VGLocations documents identifiable real-world locations represented in video games. This public preview contains a deliberately limited selection of location records for research, education, GIS, data visualization, software development, AI, machine learning, knowledge graphs, and related projects.

Dataset summary

• Release: v1.0
• Release date: 7 August 2026
• Records: 300
• Games: 265
• Franchise identifiers: 80
• Country or territory codes: 89
• Location types: 16
• Format: UTF-8 CSV with comma delimiter
• Coordinates: decimal latitude and longitude
• Split: train (300 records)

The preview is derived from the largest downloadable VGLocations dataset currently published on the project website. The source dataset contains 2,988 records across 388 games, 80 franchise identifiers, and 89 country or territory codes. The full underlying research database is not included here.

The 300 preview records were selected with a deterministic, diversity-oriented procedure that prioritizes broad representation across games, franchises, countries, and location types. The result is not a random sample and is not a complete export of the underlying research data.

One preview record does not have a country or territory code in the source data.

Dataset structure

The repository contains:

README.md
vglocations_public_preview.csv

The CSV is automatically loadable with the Hugging Face datasets library. No custom loading script is required.

Fields

Field	Type	Description
record_id	string	Stable identifier for the published record
game_title	string	Video game title associated with the location
release_year	float or missing	Release year where available
franchise	string	Franchise identifier; values correspond to the source dataset's franchise_id field
primary_genre	string	Primary game genre
secondary_genres	string or missing	Additional genres where available
location_name	string	Name of the real-world location or geographic reference
country_iso	string or missing	ISO 3166-1 alpha-2 country or territory code where available
latitude	float	Latitude of the selected geographic point, in decimal degrees
longitude	float	Longitude of the selected geographic point, in decimal degrees
location_type	string	Classification such as city, landmark, region, district, island, or battlefield
setting_status	string	Classification of the relationship between the game setting and the real-world place
coordinate_method	string	Method used to assign the coordinate
verification_level	string	Published verification classification

Loading the dataset

Replace <your-username> with the Hugging Face account or organization that owns the dataset repository.

from datasets import load_dataset

dataset = load_dataset("<your-username>/vglocations-public-preview")
print(dataset)
print(dataset["train"][0])

The CSV can also be loaded directly with pandas:

import pandas as pd

df = pd.read_csv(
    "https://huggingface.co/datasets/<your-username>/vglocations-public-preview/resolve/main/vglocations_public_preview.csv"
)

Interpreting coordinates and verification

Not every record represents a precisely surveyed point. Depending on the location, a coordinate may refer to:

• an identifiable building, landmark, or venue;
• a city, district, country, or region representative point;
• a natural feature, island, battlefield, or historical site;
• an approximate geographic anchor; or
• a real-world anchor for a fictionalized or compressed game setting.

Use coordinate_method, location_type, setting_status, and verification_level together when performing geographic analysis.

The verification_level values distinguish between verified, corroborated, and provisional records. The preview should therefore be treated as a documented research sample rather than as a claim that every record has the same evidentiary strength.

Intended uses

Appropriate uses include:

• academic and game-studies research;
• GIS and geospatial analysis;
• maps and data visualizations;
• educational projects;
• journalism and geographic storytelling;
• software and knowledge-graph development;
• internal AI, machine-learning, and retrieval experiments; and
• derived analyses and applications that do not redistribute the underlying dataset.

This dataset is not designed as a benchmark dataset, a complete catalog of video-game locations, or a uniformly geocoded corpus for high-precision surveying.

Limitations

• The dataset is a limited preview and does not represent the complete VGLocations database.
• The selection is diversity-oriented rather than statistically representative.
• Some locations are geographically compressed, fictionalized, relocated, or represented by an anchor point.
• Missing values are meaningful and should not automatically be interpreted as negative findings.
• The evidence strength is not uniform across all records.
• The dataset does not include the underlying source URLs, source identifiers, detailed evidence text, or internal research workflow fields.

License

This dataset is provided under the VGLocations Standard License, Version 1.0.

The full license terms are available in the VGLocations GitHub repository.

The license permits analysis, adaptation, research, visualization, software development, AI and machine-learning use, and other internal or derived works. It does not permit redistribution or resale of the original dataset, substantial portions of it, or a competing downloadable dataset based primarily on VGLocations data.

Attribution is strongly encouraged whenever practical:

Source: VGLocations — https://vglocations.org

Citation and related resources

Suggested citation:

Heinrich, J. (2026). VGLocations Public Preview Dataset v1.0 [Dataset]. VGLocations.

• Project website: https://vglocations.org/
• Dataset information: https://vglocations.org/video-game-locations-datasets/
• GitHub repository and release: https://github.com/theonlyduck/vglocations/releases/tag/v1.0

Contact

For corrections, questions, or suggestions:

contact@vglocations.org