Request Access to CS2 Tournament Demo Dataset
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
Please provide your affiliation and intended use of the dataset. Access requests are reviewed individually by the dataset maintainer. By requesting access, you agree to respect applicable third-party intellectual property rights and use the dataset responsibly.
Log in or Sign Up to review the conditions and access this dataset content.
CS2 Tournament Demo Dataset
The CS2 Tournament Demo Dataset is a large-scale collection of professional Counter-Strike 2 (CS2) match demo recordings sourced from HLTV.org.
The dataset covers matches from S-Tier tournaments listed on Liquipedia's Counter-Strike S-Tier Tournaments page, providing a comprehensive archive of high-level competitive gameplay.
With over 1 TB of match recordings, the dataset provides a valuable resource for research in game artificial intelligence, multi-agent learning, player behavior modeling, tactical analysis, and esports analytics.
Each match is packaged as an individual .rar archive containing its corresponding CS2 demo (.dem) files.
Dataset Overview
| Property | Description |
|---|---|
| Game | Counter-Strike 2 (CS2) |
| Source | HLTV.org |
| Tournament Coverage | S-Tier Professional Tournaments |
| Tournament Reference | Liquipedia |
| Data Size | 1+ TB |
| Data Format | Source 2 Demo (.dem) |
| Archive Format | RAR (.rar) |
| Organization | Tournament-based directories |
| Archive Granularity | One archive per match |
| Access | Gated (Manual Approval) |
Dataset Structure
The dataset is organized by tournament, with each tournament stored in a separate directory.
Each directory contains individual match archives, named according to the tournament, participating teams, match format, and an identifier.
Directory Structure
cs2dem/
├── betboom-dacha-belgrade-2024/
│ ├── betboom-dacha-belgrade-2024-aurora-vs-mibr-bo3-*.rar
│ └── ...
├── esl-pro-league-season-19/
│ ├── esl-pro-league-season-19-3dmax-vs-astralis-bo3-*.rar
│ └── ...
├── iem-cologne-2024/
│ └── *.rar
├── perfect-world-shanghai-major-2024/
│ └── *.rar
├── pgl-cs2-major-copenhagen-2024/
│ └── *.rar
├── ...
└── README.md
The directory tree is abbreviated for illustration.
File Naming Convention
Match archives generally follow the naming convention:
<tournament>-<team1>-vs-<team2>-<format>-<identifier>.rar
Example:
esl-pro-league-season-19-3dmax-vs-astralis-bo3-6rcpEL6HM941n2n6p6pXbf.rar
| Component | Description |
|---|---|
| Tournament | Tournament identifier |
| Team 1 / Team 2 | Participating teams |
| Format | Match format (BO1, BO3, BO5) |
| Identifier | Archive identifier |
Each archive corresponds to an individual match. Multi-map series may contain multiple .dem files.
Tournament Coverage
The dataset contains professional CS2 match recordings from S-Tier competitions, as classified by Liquipedia.
Covered tournament series include:
- Intel Extreme Masters (IEM)
- ESL Pro League
- BLAST Premier
- PGL Majors
- Perfect World Shanghai Major
- Esports World Cup
- Other S-Tier competitions
The collection provides match-level demo archives across major international CS2 competitions.
Dataset Access
Gated Access
Access to the dataset is restricted to approved Hugging Face users.
All access requests require manual approval by the dataset maintainer.
To request access:
- Sign in to your Hugging Face account.
- Click Request Access on the dataset page.
- Complete the access request form.
- Submit your application for review.
- Once approved, you can download the dataset.
Application Requirements
Applicants are asked to provide:
- Full name and contact information
- Institution or organization
- Country or region
- Intended use
- Research topic or project description
- Description of the planned data usage
Applications will be reviewed individually.
Access may be approved or declined at the discretion of the dataset maintainer.
Usage Conditions
By requesting access, users agree to:
- Respect applicable intellectual property rights and third-party terms.
- Use the recordings in accordance with applicable laws.
- Avoid unauthorized redistribution of the original recordings.
- Provide accurate information in their access requests.
Download
Once access has been approved, the dataset can be downloaded using the Hugging Face CLI.
Installation
pip install -U huggingface_hub
Authentication
hf auth login
Log in using a Hugging Face token associated with your approved account.
Download the Complete Dataset
hf download PriceWang/cs2dem \
--repo-type dataset \
--local-dir ./cs2dem
Download a Specific Tournament
For example, to download ESL Pro League Season 19:
hf download PriceWang/cs2dem \
--repo-type dataset \
--include "esl-pro-league-season-19/*.rar" \
--local-dir ./cs2dem
Download a Single Match
hf download PriceWang/cs2dem \
"esl-pro-league-season-19/esl-pro-league-season-19-3dmax-vs-astralis-bo3-6rcpEL6HM941n2n6p6pXbf.rar" \
--repo-type dataset \
--local-dir ./cs2dem
Extract Demo Files
RAR archives can be extracted using 7-Zip or UnRAR.
7z x match_archive.rar -o./extracted
The extracted .dem files can then be processed using compatible CS2 demo parsers.
Data Format
Counter-Strike 2 demo files use Valve's Source 2 replay format.
These recordings preserve gameplay events and player interactions.
Using compatible parsing tools, researchers can extract information such as:
- Player positions and movement trajectories
- Kills, deaths, and combat events
- Weapon usage and equipment
- Grenade and utility events
- Round and match outcomes
- Player actions and game states
- Temporal sequences of in-game events
The dataset provides raw demo recordings rather than preprocessed features, allowing researchers to develop custom data processing pipelines.
Research Applications
Multi-Agent Learning
- Multi-agent reinforcement learning
- Cooperative and competitive decision-making
- Team coordination and strategy modeling
- Offline reinforcement learning
- Imitation learning from professional gameplay
Player Behavior Modeling
- Player movement prediction
- Behavioral cloning
- Action sequence modeling
- Player decision-making
- Spatiotemporal behavior analysis
Tactical Analysis
- Team positioning and formations
- Map control and rotation analysis
- Utility usage and timing
- Tactical pattern recognition
- Strategy comparison across teams
Esports Analytics
- Professional player performance analysis
- Team strategy evaluation
- Cross-tournament comparisons
- Competitive gameplay statistics
- Match and round outcome analysis
Data Source and Attribution
The demo recordings were collected from HLTV.org.
Tournament coverage follows the S-Tier classification provided by Liquipedia.
Counter-Strike 2 is developed by Valve Corporation.
The underlying recordings and associated game content may be subject to third-party intellectual property rights. No separate license to those third-party materials is granted by this repository.
Citation
If you use this dataset in your research, please cite:
@misc{wang2026cs2tournamentdemo,
author = {PriceWang},
title = {CS2 Tournament Demo Dataset},
year = {2026},
publisher = {Hugging Face},
howpublished = {\url{https://huggingface.co/datasets/PriceWang/cs2dem}}
}
References
- Downloads last month
- 254