I work mostly on empirical asset pricing and portfolio construction: factor models, separating skill from luck in returns, dimensionality reduction on asset universes, and credit risk. Every project below starts from a question, uses public data, and keeps a strict in-sample / out-of-sample split.
| Project | Question | Methods |
|---|---|---|
| ARK Invest: Skill or Luck? | Did Cathie Wood's ETFs generate alpha, or just ride factor exposure? | Carhart 4-factor OLS, rolling regressions, bootstrap luck distribution |
| Meme-Stock Bubbles | Were GME, AMC, BB and NOK statistically in a bubble, and when? | PSY (2015) BSADF/GSADF test, Monte Carlo critical values, date stamping |
| Crypto Eigen-Portfolios | What economic forces drive the top 50 cryptocurrencies? | PCA, Marchenko–Pastur denoising, K-Means, attribution to BTC/VIX/gold/geopolitical risk |
| S&P 500 Long-Short | Can latent factor clusters drive a long-short strategy? | PCA + K-Means on factor loadings, out-of-sample backtest 2023–2025 |
| Credit Risk with Neural Networks | Which model maximises a bank's expected income on imbalanced default data? | Neural networks in R, stepwise selection, cut-off optimisation on business metrics |
| Return Analysis: NVIDIA | How do NVDA's risk and market beta evolve over time? | CAPM, rolling beta, VaR, interactive Streamlit dashboard |
Some findings
- None of the five ARK ETFs shows alpha beyond what luck explains once factor exposures are controlled for; about 80% of their monthly returns are explained by four factors.
- GameStop's 2021 episode is a clear bubble (GSADF 3.79 vs a 1.32 critical value); Nokia shows no explosive dynamics at all.
- CAPM & luck distribution: Security Market Line on STOXX 600 and Fama-French (2010) bootstrap on 300 NASDAQ stocks
- STOXX 600 factor portfolio: double sort on beta and momentum, tangency portfolios
- Growth & structural change and intertemporal choice: computational macroeconomics in MATLAB
- Italian CPI simulation: how ISTAT's NIC basket compares to a real household budget
Python (pandas, statsmodels, scikit-learn, PyPortfolioOpt, quantstats, Streamlit) · R · MATLAB · SQL (SQLite) · Excel · LaTeX