Financial Investigations

A curated collection of financial research notebooks, rendered from my GitHub repository.

Notebooks

1. Visualizing Ashwat Damodaran's analysis

Tracks Ashwat Damodaran's implied equity-risk premium, 10-year Treasury yield, analyst earnings-growth estimate, and S&P 500 level across annual, monthly, and selected daily event data. It also compares each annual ERP estimate with the S&P 500's subsequent five-year excess return.

2. Best US portfolios with ca. 100y of backtesting

Uses Damodaran's annual U.S. data since 1928 for stocks, cash, Treasury and corporate bonds, real estate, and gold. It examines rolling 3–30-year returns and drawdowns, then screens randomly generated 10%-increment portfolios by average return and worst 10-year drawdown.

3. McQuarrie's thesis on similar returns for stocks and long term bonds

Recreates Edward McQuarrie's 1793–2019 comparison of U.S. stock and long-term-bond returns in nominal and inflation-adjusted terms. It highlights the 1940–1980 bond anomaly, tests its effect on cumulative wealth, and appends recent proxy data from FRED and Yahoo Finance.

4. Extending @paolocole's comparison of market-cap, country-equally-, and stock-equally-weighted portfolios

Compares EUR-denominated MSCI market-cap- and stock-equally-weighted global indices with custom, un-rebalanced country-equal portfolios. Rolling 1-, 2-, 3-, 7-, and 12-year results from several start dates show how the return-versus-volatility comparison changes with an investor's holding period.

5. Best portfolio in USD with US, EXUS, EME stocks and Gold (1999-2025 backtesting)

Builds USD portfolios from U.S., developed ex-U.S., emerging-market equities, and gold using daily MSCI index data. A 5%-increment allocation grid is evaluated through rolling three-year annualized returns, with mean and volatility compared across multiple starting dates.

6. Recreating my own Fear and Greed Index with macro data

An in-progress market-sentiment study whose first indicator scales FINRA margin debt by the U.S. M2 money supply, then plots it alongside the S&P 500 total-return index. The notebook is intended to grow into a transparent, critiqueable set of fear-and-greed indicators.

7. Bootstrap Method 101

An incomplete stationary block-bootstrap prototype for SPY, VT, GLD, and IEF that preserves cross-asset correlation and short-run return dependence. It simulates 5-, 10-, 20-, and 30-year outcomes for every 10%-increment allocation and compares median and downside annualized returns.

Legacy

This repository was previously named "Saturday Morning Projects" to experiment with copilot + fun research.

Then several projects moved to dedicated repositories, and this one was renamed to "Financial Investigations" to focus on financial research.

I stated the following rules for the notebooks:

State the AIM of the analysis at the beginning of the project
End the notebook with a Conclusions section, summarizing the findings, and a Follow-up section, listing possible next steps to take
Make intense use of GitHub Copilot
Explore creative uses of plotly for interactive visualizations, and remember to use nbmirror properly
Don't spend more than 2 hours on a single notebook - this is a Saturday morning activity, not a full-blown project that will take weeks to complete
The previous point was totally suggested by Copilot... I'm not sure if I agree with it, but it's a fair reminder that after 2h you get stuck in recent opinions and it is better to stop and come back later
Don't be too strict in using statistically sound methods (for sake of time), but be open to critics in case the analysis was too handwaving
Come back from time to time, to revise the narrative of the analysis to make it clearer and clearer
Be totally open to comments, suggestions, and collaborations: for any of them, open an Issue or a Pull Request
Don't have more than 10 rules