PortfolioIQ — Equity Risk & Returns Engine

Inspiration

As mathematics students, we wanted to bridge the gap between theoretical finance and real market data. Managing a ₹10,00,000 hypothetical portfolio across 15 NSE stocks pushed us to think like actual quant analysts.

What We Built

A complete equity analysis pipeline covering:

Task 1: Fetched 2 years of daily OHLCV data (Jan 2023 – Dec 2024) for 15 NSE stocks using yfinance
Task 2: Computed 6 risk metrics per stock — Annualised Return, Volatility, Sharpe Ratio, Beta vs Nifty 50, Max Drawdown, SMA Signals
Task 3: Technical signal dashboard with Golden Cross / Death Cross detection using 50-day and 200-day SMA
Task 4: Constructed Portfolio A (equal weight) vs Portfolio B (Sharpe-optimised) with sector breakdown
Task 5: Automated live price alert via n8n → Telegram Bot
Task 6: ML classifier (Random Forest vs Gradient Boosting) to predict relative outperformance

Key Results

Metric	Value
Best ML Model	Gradient Boosting (56.3% accuracy)
Top ML Feature	RSI_14 (importance: 0.314)
Best Sharpe Stock	SUNPHARMA (1.94)
Most Exposed (Beta)	SBIN (β = 1.394)
Safest Refuge	CIPLA (β = 0.333)

Stress Test (Chaos Round)

If Nifty 50 drops 10%:

$$\text{Expected Loss} = \beta \times 10\%$$

Portfolio	Weighted Beta	Expected Loss
Portfolio A (Equal Weight)	0.799	-7.99%
Portfolio B (Recommended)	0.538	-5.38%

Portfolio B loses 2.6% less than Portfolio A in a market crash.

Challenges

Handling missing trading days and NSE market holidays in raw data
Avoiding data leakage in the ML temporal train/test split
Building a live n8n workflow that fetches and evaluates 3 stocks in a single Code node without external dependencies