Skip to content

The crucible Podcast

An audio companion to the tutorial — the same argument, out loud: how a rising equity curve becomes a defensible verdict, and why so many "edges" don't survive the question.

⬇️ Download the episode (M4A) · 📖 Read the full tutorial · 📄 Tutorial as a PDF

What the episode covers

The tutorial's arc, narrated end to end:

  • Luck vs. skill, and patterns that don't travel — the two failure modes every headline backtest number hides, and why a point estimate isn't a verdict.
  • The capital-free scorecard — expectancy, profit factor, payoff, SQN, and excursion efficiency, all before any position sizing or equity curve.
  • Quantifying the noise — bootstrap confidence intervals and p-values that turn "profit factor 1.34" into a claim you can defend.
  • Ruling out data-mining luck — sign-permutation tests and the random-entry reality check: could no edge at all have produced this?
  • The gauntlet — REAL → STRONG → DURABLE → GENERAL, run as one ordered set of hard gates that returns a single audited pass/fail.

Prefer to read?

Everything discussed here is worked out in full, with code you can run (pip install crucible) and references to the source literature, in the tutorial.