TQBTHEQUANTBATEMAN
TQB/ deskEN · DARK
THE DESK · EDITORIAL PROTOTYPE

NOTES WITH
CONSEQUENCES.

Models, markets and research—written for people who eventually have to hedge the answer.

FEATURED · MODEL DEBATES
σloc(S,t)

Is Local Vol Actually Useful?

Perfect vanilla calibration, deterministic smile dynamics, and a very particular relationship with forward skew.

A fit is a property of today’s surface. A hedge is a theory about tomorrow’s.
QUANT THOUGHTS02

Sticky Delta vs Sticky Strike

What are we actually assuming when spot moves and the implied-volatility surface must follow—or refuse?

Your surface moved. The convention merely chose how to describe the scene.
RESEARCH03

Why your model calibration is not your model.

Objective functions, identifiability, parameter stability and the things a small RMSE does not tell you.

Excellent. The optimizer has given up before you did.
QUESTIONS04

Should every desk own a differentiable pricer?

AAD, automatic differentiation and the architecture tax behind attractive gradient benchmarks.

A gradient is only cheap after somebody pays for the tape.