Trader Sentiment Analysis

An exploratory data analysis project that studies how Bitcoin Fear & Greed sentiment relates to trader behavior and performance on Hyperliquid trading data.

What This Demonstrates

• Data cleaning and feature engineering
• Daily account-level aggregation
• Sentiment-regime comparison
• Trader segmentation by activity level
• Strategy recommendations from observed behavior
• Lightweight Streamlit dashboard

Tech Stack

• Python
• Pandas
• NumPy
• Matplotlib
• Seaborn
• Streamlit

Analysis Flow

1. Load Fear & Greed index data and historical trader execution data.
2. Normalize dates and join each trade to the sentiment regime for that day.
3. Create behavior and performance features:
   • daily PnL
   • win rate
   • average trade size
   • trades per day
   • long/short bias
4. Compare Fear and Greed periods.
5. Segment traders by frequency.
6. Generate strategy signals for risk-aware behavior changes.

Key Findings

• Greed days show higher upside but also higher variability.
• Fear days tend to produce more cautious behavior and lower activity.
• Frequent traders are more sensitive to sentiment shifts.
• Position sizing and trade frequency are important risk controls across regimes.

Run Locally

pip install -r requirements.txt
jupyter notebook notebooks/analysis.ipynb

Run the dashboard:

streamlit run dashboard.py

Outputs

outputs/tables/performance_summary.csv
outputs/tables/dashboard_data.csv

Portfolio Note

This project is analysis-focused rather than production software. The next upgrade would be adding reproducible pipeline scripts, validation checks, and a clearer model evaluation section.