FragAudit parses CS2 .dem files and generates tactical analysis:
- Mistake Detection — Flags overpeeks, failed trades, spacing errors
- Role Classification — Entry, Lurk, Anchor, Support per round
- Win Probability — Round-by-round predictions with factor breakdown
- Strategy Detection — Execute, Rush, Split, Default patterns
- Radar Replay — MP4 video with player positions and utility
# Install
git clone https://github.com/Pl4yer-ONE/FragAudit.git
cd FragAudit
python -m venv venv && source venv/bin/activate
pip install -r requirements.txt
# Verify
python -m pytest tests/ -q # 174 passed
# Analyze
python main.py analyze --demo match.dem --html --radar| Flag | Speed | Quality | Use Case |
|---|---|---|---|
--radar |
5-6 min | High (view cones, effects) | Final review |
--radar --fast-radar |
90 sec | Standard | Quick preview |
# Fast radar (3.7x faster)
python main.py analyze --demo match.dem --radar --fast-radar| File | Description |
|---|---|
reports/coaching_report_*.json |
Raw analysis data |
reports/report_*.html |
Visual HTML report |
reports/radar_*.mp4 |
Radar replay video |
from src.predict import predict_round_win
result = predict_round_win(
team_economy=4500,
enemy_economy=4500,
team_alive=5,
enemy_alive=3 # 5v3 advantage
)
print(result.probability) # 61.8%
print(result.dominant_factor) # man_advantage| Module | Description |
|---|---|
src/mistakes/ |
5 error types with severity |
src/roles/ |
Entry, Lurk, Anchor, Rotator, Support |
src/wpa/ |
Contextual win probability |
src/strategy/ |
Strategy pattern detection |
src/predict/ |
Win/impact prediction |
src/radar/ |
Radar video generation |
| Operation | Time |
|---|---|
| Demo analysis | ~7s |
| Radar (standard) | ~6 min |
| Radar (fast) | ~90s |
| Unit tests | 0.8s |
HTML Report
Player Cards
GNU General Public License v3.0
Radar maps from boltobserv (GPL-3).
See THIRDPARTY.md for attribution.
174 tests. No ML libraries. Explicit coefficients.