Weekly data/ML and finance projects in Python. An incubator for new builds.
- Stock Data Toolkit — pull, analyze, and chart historical stock data (returns, volatility, drawdown, moving averages).
- Portfolio Lab — multi-asset portfolio analysis with correlation and mean-variance optimization: closed-form and long-only constrained min-variance & max-Sharpe portfolios, plus the efficient frontier.
- Backtester — simulate holding a portfolio through time with periodic rebalancing and transaction costs, plus point-in-time signal strategies (momentum, inverse-volatility) with no lookahead; equity-curve and drawdown charts. A three-part honesty toolkit: no-lookahead, walk-forward validation with an "overfitting tax," and block-bootstrap significance testing (is the Sharpe distinguishable from luck?). Graduated to its own repo → GalacticChill/honest-backtester (pip-installable, CI-tested).
- Star Signals — does GitHub reward substance or surface signal? A machine-learning capstone predicting a repo's stars from features split into signal vs substance, measuring how much predictive power each camp contributes. Graduated to its own repo → GalacticChill/star-signals (pip-installable, CI-tested).
- Market Structure — unsupervised learning on daily returns: can hierarchical clustering rediscover the market's sectors from price co-movement alone? Dendrogram, reordered correlation heatmap, and an honest adjusted-Rand score against real sectors.