Series
Edge, sizing, validation, and the closed loop
A long-form series drawn from the Financial AI textbook — probability and EV, Kelly sizing, signal discovery, microstructure, validation gauntlets, regimes, LLM governance, and daily alpha closure.
Start at Ep00 (how to read), then Part I: the biased coin and EV. Each part adds an angle on the same object — sizing, discovery, combination, microstructure, validation, regime, LLMs, then synthesis.
Payoff chapters: Ep36 (convergence card), Ep56 (Daily Alpha Closure), Ep58 (day in the life). Companion textbook: Financial AI by Srikanth Chandragiri.
Edge is a property of a class. EV, LLN, CLT, and risk of ruin.
Believe in distributions. Kelly, Brier, drawdowns, deployment.
Read after Part I.
IC is tiny. Test honestly — Bonferroni, HAC, effective N.
Read after Part I.
Incremental information via regression, ACE, and probability scores.
Read after Part I.
Friction, flow, stops, halts, impact, concentration.
Read after Part I.
Trust gauntlet → convergence card → labels → attribution.
Read after Part I.
Environment is not i.i.d. Size and execute in context.
Read after Part I.
Mechanism, alignment, sampling, tools, evaluation, promotion governance, and the domain tax.
Read after Part I.
Daily Alpha Closure, behaviour, narrative day, two-loop improvement.
Read after Part I.
Aliases and fleets, priority admission, the verify pass, training-velocity economics, compile-first memory, and agent-org design — the software layer, not the mathematics.
Read after Part I.