Rough volatility · interactive tour · github.com/Michaellumor/roughvollab

Inside RoughVolLab

A single tested core, and five honest questions about how rough markets really are — can you identify it? can you exploit it? Fact, or artefact? Each card below opens a hands-on explorer you can drive yourself.

Family A · can you identify it?
Q1 identify · from data live
Can you measure roughness back?
Three standard estimators try to recover the true H — and disagree most exactly where real markets live.
— often, noOpen the explorer →
Q4 identify · from option prices live
Can you back it out of the smile?
Calibrate rough-Heston to an option surface. A single smile can’t pin H down; on live Deribit it rails to the boundary.
— noOpen the explorer →
the foundation · layer 1
The trusted engine
roughvol_core.py
The κ=0 Volterra core every layer is built on. Drive it, flip on the old normalisation bug, and watch the test catch what the eye can’t.
Open the engine →
Family B · can you exploit it?
Q3 exploit · trading live
Can you trade against it?
Time execution to the vol and race the classic Almgren–Chriss schedule. On matched risk it lands ≈5σ worse — the first “edge” was a look-ahead bug.
— noOpen the explorer →
Q5 exploit · hedging live
Can you hedge with it?
Deep hedging beats the delta hedge under costs — but that’s a generic frictions effect. The roughness-specific gain is absent (z = 1.4).
— noOpen the explorer →
Q2 the bridge · pricing live
Can you price it cheaply?
Pricing options under rough vol needs heavy Monte Carlo. A conditional single-grid estimator beats the fashionable multilevel MC outright — a decisive, κ-invariant cost ratio of 0.41–0.45.
— yes, qualifiedOpen the explorer →
6 of 6 explorers live
identify exploit the pricing bridge
Open-source quantitative research · honest measurement over confident numbers
M. Lumor · University of Salford
How this codebase audits itself →
github.com/Michaellumor/roughvollab