Why parametric assumptions blind your models—and how distribution-free methods keep you adaptive
Markets don’t follow tidy distributions. And every time you hardcode a normality assumption, you’re gambling with fragility. The solution? Let the data speak for itself.
This chapter gives you the tools to model uncertainty without relying on brittle parametric guesses.
What’s inside:
🔹 Inference without assumptions: Rank-based methods, U-statistics, and permutation tests for robust signal extraction without a distributional crutch.
🔹 Bootstrap like a quant: Resample edge, risk, and drawdowns with empirical confidence—and simulate what traditional models miss.
🔹 Jackknife mastery: Estimate stability, bias, and influence—critical for evaluating fragile strategies and noisy alpha sources.
🔹 Conformal prediction decoded: Learn how this game-changing method builds distribution-free prediction intervals—perfect for live systems with changing regimes.
🔹 Python tools included: End-to-end scripts for resampling-based inference, conformal quantiles, and signal calibration—all tested on real market data.
🔹 Quantitative power, unchained: Strip away the Gaussian myth—model tail risk, dependency, and uncertainty the way markets actually behave.
This isn’t stats for clean data. This is inference for chaos—tools built to survive the dirty, unpredictable, and distributionless real world of trading.