Hugging Face's logo

Datasets:
KEDevO
/
CounterQuant-CS2-Demos

Tasks:
Other
Languages:
English
Size:
n>1T
Tags:
cs2
counter-strike
esports
demos
sports-analytics
game-theory
License:
Dataset card Files
xet
 Community
CounterQuant-CS2-Demos / README.md
Unit293's picture
Unit293
Update dataset card: Eimantas Kulbe credit, coverage plan T1/T2/T3, usage examples
c9d63e3 verified
preview code
|
raw
history blame
4 kB
metadata 
license: cc-by-4.0
task_categories:
  - other
language:
  - en
tags:
  - cs2
  - counter-strike
  - esports
  - demos
  - game-theory
  - sports-analytics
pretty_name: CounterQuant CS2 Demos
size_categories:
  - 10K<n<100K

CounterQuant CS2 Demos

Raw CS2 (Counter-Strike 2) demo files (.dem) collected from HLTV — the authoritative competitive CS2 demo repository.

Curated and maintained by Eimantas Kulbe as part of the CounterQuant esports analytics project.

Dataset Goal

Provide the CS2 research community with the largest open-access collection of raw professional demo files, covering all competitive tiers from CS:GO's origin through CS2's present day.

The philosophy: squeeze every available demo into one place so anyone can preprocess, parse, or derive features in any way they want — no gatekeeping, no preprocessing lock-in.

Coverage

Tier Description Years Status
T1 Major/S-tier events (top 20 world teams) 2024–present Active — daily updates
T2 A-tier events (top 50 teams) 2024–present Active — daily updates
T3 B-tier / regional events 2024–present In progress
Historical CS:GO T1/T2 (pre-2024) 2012–2023 Planned

File Structure 

data/
  {year}/
    tier{N}/
      {match_id}/
        {map_name}.dem

Example:

data/2024/tier1/2378549/astralis-vs-natus-vincere-mirage.dem
data/2025/tier2/2401337/team-vitality-vs-g2-inferno.dem

Usage

Parse with demoparser2 (Python) 

from demoparser2 import DemoParser

parser = DemoParser("data/2024/tier1/2378549/astralis-vs-natus-vincere.dem")

# Player kill events
kills = parser.parse_event("player_death", player=["name", "team_name"])

# Round-by-round economy
rounds = parser.parse_ticks(["total_rounds_played", "cash_spent_t", "cash_spent_ct"])

Download via huggingface_hub 

from huggingface_hub import snapshot_download

# Download all 2024 T1 demos
path = snapshot_download(
    repo_id="KEDevO/CounterQuant-CS2-Demos",
    repo_type="dataset",
    allow_patterns="data/2024/tier1/**",
    local_dir="./demos",
)

Stream individual files 

from huggingface_hub import hf_hub_download

dem = hf_hub_download(
    repo_id="KEDevO/CounterQuant-CS2-Demos",
    repo_type="dataset",
    filename="data/2024/tier1/2378549/match.dem",
)

What You Can Do With These Demos

  • Feature engineering: extract round economy, positioning, utility, player rating
  • Model training: win-probability, player rating, team strength models
  • Research: CS2 game theory, tactical analysis, team dynamics
  • Statistics: compute your own KDA, HLTV Rating 2.0, ADR, Impact
  • No restrictions: raw bytes, parse however you like

Update Schedule

New demos are added within 24–48 hours of match completion. The pipeline runs continuously across dedicated VPS infrastructure.

Data Source

Demos sourced from HLTV.org — the official repository for professional CS2 match data. HLTV is the authoritative source for all competitive CS2/CS:GO results and replays.

Citation

If you use this dataset in research, please cite:

@dataset{kulbe2026counterquant,
  author    = {Eimantas Kulbe},
  title     = {CounterQuant CS2 Demos},
  year      = {2026},
  publisher = {HuggingFace},
  url       = {https://huggingface.co/datasets/KEDevO/CounterQuant-CS2-Demos},
  note      = {Continuously updated professional CS2 demo dataset, all tiers, T1/T2/T3}
}

License

CC BY 4.0 — use freely with attribution to Eimantas Kulbe / CounterQuant.

Related Resources