Skip to main content

Hybrid & Practical Models

Pure models are clean but limited. Stochastic vol gets the dynamics right but not the smile. Local vol gets the smile right but not the dynamics. These hybrids combine both -- and are what production desks run.

💡
SLV is the industry default for exotics

Ask a quant what model they use for exotic pricing. The answer is almost always stochastic local vol -- local vol combined with a stochastic vol layer.

At a Glance

Model
What it combines
Best for
<a href="/docs/reference/stochastic-local-vol">Stochastic Local Vol (SLV)</a>
Local vol + stochastic vol
Production exotic pricing. The industry workhorse.
<a href="/docs/reference/vanna-volga">Vanna-Volga</a>
Black-Scholes + 3-quote adjustment
Quick FX smile construction from ATM, RR, BF quotes.

What they share

Both models combine simpler components to solve problems that pure stochastic vol or pure local vol cannot handle alone. They differ in complexity, inputs, and target use case.

Model
Inputs needed
Calibration
Exotic pricing?
Speed
SLV
Full vol surface + stochastic vol params
Particle method or PDE (heavy)
Yes
Slow
Vanna-Volga
3 quotes: ATM, risk reversal, butterfly
Analytic (closed-form adjustment)
Limited
Very fast

How they relate to each other

Stochastic Local Vol is the production workhorse for exotic pricing. It takes a local vol surface (calibrated to match all vanilla prices exactly) and overlays a stochastic vol process, giving it realistic dynamics for path-dependent products. The cost is calibration complexity -- it requires particle methods or PDE solvers and a full vol surface as input. Vanna-Volga sits at the other end of the spectrum. It constructs a smile from just three FX market quotes (ATM, risk reversal, butterfly) by computing the cost of hedging vanna and volga risk with those three instruments. It is fast, analytic, and widely used on FX desks, but it does not extend naturally to exotic pricing or non-FX markets.


Models in this section: