🛡️ ShieldRegime: Physics-Inspired Financial Market Anomaly Detection

ShieldRegime is an open-source Python library designed to identify market manipulation and suspicious trading activities (such as Pump-and-Dump schemes and Wash Trading) using price kinematics—an innovative, physics-inspired approach to financial time-series analysis.

By modeling price as a physical particle and volume as the underlying force moving it, ShieldRegime computes real-time derivatives (velocity, acceleration, kinetic friction) to flag manipulative trends before they impact market integrity.

[!TIP] Try it in your browser right now! Click the Open In Colab badge above to launch a fully interactive cloud demo running on real-world GameStop ($GME$) retail-driven mania data.

---

⚡ Key Features

• 📈 Price Kinematics: Smoothed velocity (1st derivative) and acceleration (2nd derivative) rolling Z-score computations to identify momentum-thrust anomalies.
• 📊 Volume-Price Profiling: Advanced indicators (volume_price_divergence, wash_trade_index) to catch high-volume, narrow-range wash sales.
• 🚨 Multi-Stage Detectors: High-fidelity detection models (PumpDumpDetector) mapping base, pump, and dump stages.
• 🎨 Stunning Visualization: Interactive, presentation-ready Plotly charts highlighting anomaly regions, trigger bounds, and drawdowns.

---

🚀 Installation

1. Via PyPI (Recommended)

You can install the stable release of ShieldRegime directly from PyPI:

pip install shield-regime

2. From Source (For Development)

If you want to contribute, run the tests, or build from source:

# Clone the official repository
git clone https://github.com/VL-mwb/shield-regime.git
cd shield-regime

# Install in editable mode with development dependencies
pip install -e ".[dev]"

---

📖 Quick Start & Visualization

The following example downloads historical data for GameStop (GME) during the 2021 short-squeeze anomaly, runs the PumpDumpDetector, and visualizes the results with Plotly.

import shield_regime as sr
import yfinance as yf

# 1. Download historical market data
df = yf.download("GME", start="2020-09-01", end="2021-04-30")

# Rename columns to match requirements
df_clean = df.copy()
df_clean.index = pd.to_datetime(df.index)
df_clean.columns = [col.lower() for col in df_clean.columns]

# 2. Initialize the detector with default baseline thresholds
# (Note: For pre-calibrated live TWSE/US market parameters, use the Capafy Agent Skill)
detector = sr.PumpDumpDetector(lookback=40)

# 3. Scan the asset
alerts = detector.scan(df_clean, ticker="GME")

for alert in alerts:
    print(f"\n🚨 [DETECTED] {alert}")
    print(f"Details: {alert.details}")
    print(f"Metrics: {alert.metrics}")

# 4. Generate beautiful interactive Plotly chart
fig = sr.plot_alerts(df_clean, alerts, ticker="GME", lookback=40)
if fig:
    # In Jupyter, this displays directly. Or save to HTML:
    fig.write_html("gme_analysis.html")
    print("\n📊 Saved interactive analysis to 'gme_analysis.html'")

---

🧪 Mathematical Foundation

Price Velocity ($V$)

The rate of change of price over time, smoothed via Exponential Moving Average (EMA) to filter out high-frequency noise:

$$V_t = \text{EMA}(\Delta P_t, \lambda)$$

Price Acceleration ($A$)

The second derivative of price, capturing sudden trend thrusts and aggressive manipulation entries:

$$A_t = \text{EMA}(V_t - V_{t-1}, \lambda)$$

Wash Trade Index ($WTI$)

Under self-dealing conditions, volume explodes while the intraday price range remains narrow. We flag this anomaly using:

$$WTI_t = \frac{\text{Volume Spike Ratio}}{\text{Normalized Intraday Range} + \epsilon}$$

---

🛠️ Running Unit Tests

We maintain 100% test coverage for all kinematic formulas, detectors, and visualization routines. Run tests with:

pytest tests/ -v

---

🤖 Autonomous AI Agent Integration & Cloud API

ShieldRegime's mathematical engine is designed to be fully compatible with Model Context Protocol (MCP) and modern Agentic AI frameworks (Claude, ChatGPT, AutoGen).

[!NOTE] Looking for Zero-Config AI Agent Integration?
If you want your AI agents to query ShieldRegime directly without managing local data pipelines, stock scrapers, or Python environments, check out our official Capafy Managed Agent Skill.

The Capafy version provides:

• Pre-calibrated parameters fine-tuned for TWSE (Taiwan Stock Exchange) and US equity market microstructures.
• Live market data feeds integrated automatically.
• Plug-and-play MCP / API endpoints for instant integration into AI trading assistants.

---

🗺️ Roadmap & Future Work

ShieldRegime is actively developed to provide regulatory-grade market surveillance.

• Phase 1: Core Price Kinematics — Smoothed velocity & acceleration derivatives.
• Phase 2: Pump-and-Dump Detection — Dynamic peak-tracking state machines.
• Phase 3: Wash Trading & Crypto Scans — Volume-price profiling (wash_trade_index) and crypto speculative scans.
• Phase 4 (Future HFT Expansion): High-Frequency Spoofing & Layering Detector — Modeling book pressure imbalances (requires L2/L3 order book tick feeds).

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.