← All projects Independent research
Quantitative Portfolio Optimization
I compared ways to allocate a portfolio and tested how the results change when market assumptions become less favorable.
Independent Project · 2025 - August 2026
Why I built it
A portfolio allocation can look sensible under one set of assumptions and change sharply under another. I wanted to compare methods and understand how much their results depend on expected returns, risk estimates and constraints.
What I implemented
- Portfolio comparisons using Markowitz, Black-Litterman, Risk Parity and CVaR methods.
- Scenario analysis for changing volatility, correlation, transaction costs and downside risk.
- Monte Carlo-style tests and sensitivity analysis to examine how stable the allocations were.
How I used the comparisons
I changed the inputs and reviewed how the weights, concentration and risk measures responded. That made it possible to compare each method's assumptions and failure modes alongside its historical output.
What I can demonstrate
The project produced implemented comparisons across several allocation approaches and a way to examine their sensitivity to market assumptions. The useful outcome was understanding when an allocation becomes unstable and which inputs drive that behavior.
Python supports the allocation comparisons, simulation and scenario analysis.
PythonPortfolio allocationMonte CarloScenario analysis
Engineering details
Open a topic for the implementation decisions and checks behind the overview.
What the allocation methods optimize
Markowitz balances expected return against variance. Black-Litterman combines a prior with explicit views. Risk Parity focuses on each holding's contribution to risk. Conditional Value at Risk (CVaR) examines losses in the tail.
Monte Carlo and scenario analysis test sensitivity to changing estimates and constraints. The comparison also considers volatility, covariance, drawdown and transaction-cost assumptions.
Current scope
This is independent research. Reproducible comparison artifacts are not published here, and the portfolio makes no audited-return, method-ranking or production-alpha claim.