📈 Return Analysis

A collection of quantitative equity analyses built in Python — one stock at a time.
Each analysis combines a Jupyter notebook for research and a Streamlit dashboard for interactive visualization.

---

What this repo is about

This project applies standard quantitative finance tools to individual equities, starting with NVIDIA (NVDA).
The goal is to build a repeatable, modular framework that can be extended to any publicly traded stock.

For each company, the workflow is:

1. Jupyter Notebook — downloads price data via yfinance and Fama-French factors, computes excess returns, runs descriptive statistics, fits a CAPM regression via OLS, and exports results to Excel.
2. Excel file — the output of the notebook, used as the data source for the dashboard (no live internet connection required at runtime).
3. Streamlit Dashboard — loads the pre-computed Excel, renders interactive charts and performance metrics using Plotly.

---

Current analyses

Company	Ticker	Period	Notebook	Dashboard
NVIDIA Corporation	NVDA	Jan 2016 – Feb 2026	NVIDIA/NVIDIA.ipynb	🚀 Launch Dashboard

Dashboard overview: NVDA vs the US market over the full period.

What the dashboard shows

• Performance overview: CAGR, volatility, Sharpe, Sortino, max drawdown, VaR and CVaR, each compared with the market
• Growth of $1 and drawdowns: NVDA vs the market, linear or log scale
• Calendar-year returns
• Rolling risk: annualised volatility and CAPM beta on a 3-month, 6-month or 1-year window
• Return distribution: histogram vs normal, Q-Q plot, moments and Jarque-Bera test
• CAPM regression: scatter with the characteristic line, OLS table, annualised alpha and a plain-language reading of the results

Every number updates with the selected period (1Y, 3Y, 5Y, Max).

From the notebook: daily NVDA excess returns vs market excess returns, with the fitted CAPM line.

---

Methodology

Data

• Prices: adjusted close prices from Yahoo Finance via yfinance
• Risk-free rate & market factor: daily Fama-French factors (Mkt-RF, RF) from Kenneth French's data library

Metrics computed

• Distributional moments: mean, standard deviation, skewness, excess kurtosis
• Performance: CAGR, annualized volatility, Sharpe ratio, Sortino ratio
• Risk: maximum drawdown, historical VaR and CVaR at 95% (daily)
• CAPM regression (OLS): alpha (daily and annualised), beta, R², t-statistics, p-values, Durbin-Watson
• Rolling volatility and beta: moving-window estimates to capture time-varying risk

The notebook uses quantstats. The dashboard computes the same metrics directly with NumPy so that it can recompute them for any period; VaR and CVaR are historical, and CAGR and drawdowns use total returns (excess return + risk-free rate).

---

Repository structure

Return_Analysis/
└── NVIDIA/
    ├── NVIDIA.ipynb          # Research notebook
    ├── nvda_analysis.xlsx    # Output of the notebook (data source for dashboard)
    ├── requirements.txt      # Notebook dependencies
    └── dashboard/
        ├── nvda_dashboard.py # Streamlit app
        ├── nvda_analysis.xlsx
        └── requirements.txt

---

How to run

1 — Notebook

Open NVIDIA/NVIDIA.ipynb in Jupyter and run all cells.
This will generate nvda_analysis.xlsx with four sheets: NVDA_Prices, FamaFrench_3F, ER_Analysis, OLS_Results.

2 — Dashboard (local)

pip install -r NVIDIA/dashboard/requirements.txt
streamlit run NVIDIA/dashboard/nvda_dashboard.py

2 — Dashboard (Streamlit Cloud)

The app is deployed on Streamlit Cloud. Dependencies are declared in requirements.txt and picked up automatically.
The dashboard reads nvda_analysis.xlsx from its own folder — no live data download at runtime.

---

Stack

Python · pandas · numpy · yfinance · statsmodels · quantstats (notebook) · scipy · plotly · streamlit · openpyxl

---

Roadmap

• Apple (AAPL)
• Microsoft (MSFT)
• Tesla (TSLA)
• Multi-stock comparison notebook
• Fama-French 3-factor and 5-factor extensions

---

Author

Feel free to open an issue or a pull request if you want to contribute an analysis for another stock.