Convex Risk Notebook

Companion notebook for the post “Algo Backtests Are Lying to You.” It recreates the article’s synthetic experiments, then reruns the same ideas on live SPY and BTC-USD candles via backtesting.py.

Quick start

• Install deps with uv sync (or uv pip install -r pyproject.toml).
• Launch Jupyter: uv run --frozen jupyter lab (or uv run --frozen jupyter notebook).
• Open python/convex.ipynb and run cells top-down. Section 6 fetches real data; you’ll need network access for the yfinance pulls.

Notebook map

1. Gaussian comfort zone — deterministic SMA edges in a tidy world.
2. Fat tails, shocks, quantile payoffs, and Monte Carlo convex/concave contrasts.
3. Regime-shift stress test synthesizer.
4. Real-data sanity check with backtesting.py across SPY and BTC regimes.

Feel free to adjust the RNG seed, regime definitions, or add your own instruments to falsify strategies faster.