Interpolation Methods for Vol Surfaces

info

This is a companion page to How Vol Surfaces Are Built. Start there for context on why interpolation matters.

The vol surface has gaps. Interpolation fills them. The choice of method determines whether the resulting surface is smooth, arbitrage-free, and stable. This page compares the major approaches.

Interpolation Methods Compared

The white dots are the only real market quotes. Everything in between is estimated. Click each method to see its strengths and weaknesses.

What Goes Wrong

Before diving into each method, see the problems for yourself. The same 7 market observations, three different interpolation methods. Watch what happens in the wings and at the data points.

What Goes Wrong: Interpolation Failures

Same 7 market observations, three different interpolation methods. Watch what happens in the wings.

Smooth polynomial curves. Can oscillate and overshoot in the wings.

White dots are the only real observations. Click "Compare All" to overlay all three methods. Notice how the spline overshoots in the left wing while SVI stays bounded.

---

Non-Parametric Methods

These methods fit curves through data points without assuming a functional form. They are fast and simple, but they offer no structural guarantees.

Linear interpolation

Draw straight lines between adjacent data points.

Advantages:

• Trivial to implement
• No fitting or optimization
• Deterministic: same inputs always give same outputs

Disadvantages:

• Creates sharp corners at every data point. These corners produce discontinuous first derivatives, which means Greeks (especially gamma) jump abruptly at observed strikes.
• No guarantee against butterfly arbitrage. A straight line between two points can dip below where a convex smile should be.
• Extrapolation is pure speculation (just extends the last segment's slope).

Use it for: Quick estimates, sanity checks, debugging. Not production pricing.

Cubic spline interpolation

Fit piecewise cubic polynomials between data points, constrained so the first and second derivatives match at each joint. The result is a $C^2$-smooth curve (continuous curvature).

The name comes from physical drafting splines: flexible strips of wood that draftsmen bent through pins to draw smooth curves.

Advantages:

• Smooth curve through all data points
• No parameter estimation (the spline is determined by the data and boundary conditions)
• Fast to compute

Disadvantages:

• Runge's phenomenon: At the edges of the interpolation domain, the polynomial can overshoot wildly. For vol surfaces this means wing IVs that spike or go negative.
• Oscillation: Between data points, the cubic can swing above or below what a well-behaved smile would produce, creating concave dips (butterfly arbitrage).
• Sensitivity to outliers: One bad data point (stale quote, fat finger) distorts the entire curve because the smoothness constraints propagate the error.
• No control over extrapolation behavior.

Use it for: Visualization, academic work, or as an initial guess before parametric fitting. Not for production pricing or risk.

---

Parametric Methods

These methods assume a functional form for the smile and fit its parameters to the data. They trade exact interpolation for structural control.

SVI (Stochastic Volatility Inspired)

The industry standard for crypto and equity vol surfaces. Five parameters per expiry slice.

$$w(k) = a + b \left( \rho(k - m) + \sqrt{(k - m)^2 + \sigma^2} \right)$$

See the full reference: SVI Parameterization

Why it dominates: SVI is the sweet spot between flexibility and parsimony. Five parameters can fit nearly any observed smile shape, while simple inequality constraints guarantee no butterfly arbitrage. The wings approach linear asymptotes, so extrapolation is bounded and sensible.

SABR (Stochastic Alpha Beta Rho)

A stochastic volatility model that derives the smile from assumptions about how vol evolves. Four parameters: $\alpha$ (vol level), $\beta$ (CEV exponent), $\rho$ (spot-vol correlation), $\nu$ (vol-of-vol).

See the full reference: SABR Model

Why it exists: SABR captures the dynamics of the smile, not just the static shape. It tells you how the smile should move when the underlying moves (sticky delta by default). This makes it natural for interest rate swaptions where the smile dynamics matter for hedging.

Local Volatility (Dupire)

Not a fitting method in the usual sense. Local vol derives an instantaneous volatility surface from the observed implied vol surface. It answers: "What must the instantaneous volatility be at each (spot, time) combination to reproduce these option prices exactly?"

See the full reference: Local Volatility

Why it exists: Local vol is the unique arbitrage-free model that exactly matches all observed option prices. It is the bridge between implied vol and a pricing engine that can handle path-dependent payoffs.

SSVI (Surface SVI)

An extension of SVI that models the entire surface jointly, not slice by slice. SSVI enforces calendar arbitrage freedom by construction: total variance is guaranteed to increase with maturity at every strike.

$$w(k, \theta_t) = \frac{\theta_t}{2} \left( 1 + \rho \, \varphi(\theta_t) \, k + \sqrt{(\varphi(\theta_t) \, k + \rho)^2 + (1 - \rho^2)} \right)$$

Where $\theta_t$ is ATM total variance at time $t$ and $\varphi(\theta_t)$ controls how the skew evolves with maturity.

Trade-off: Fewer free parameters than per-slice SVI (the smile shape is linked across expiries), so the fit may be slightly worse at individual slices. But you never need post-hoc calendar arbitrage fixups.

---

Comparison Table

Method

Parameters

Arb-Free?

Extrapolation

Speed

Best For

Linear

0

No

Unbounded

Instant

Debugging

Cubic Spline

~12 (implicit)

No

Oscillates

Fast

Visualization

SVI

5 per slice

Yes (constrained)

Bounded linear

Fast

Crypto / equity

SABR

4

Mostly

Reasonable

Medium

Rates / swaptions

Local Vol

Full grid

By construction

N/A (derived)

Slow

Exotic pricing

SSVI

~6 (surface)

Yes (calendar too)

Bounded

Fast

Full surface consistency

How to choose

• For production crypto/equity pricing: SVI or SSVI. The industry has converged here for good reason.
• For interest rate options: SABR. It captures smile dynamics that matter for swaption hedging.
• For exotic derivatives pricing: Local vol (or a stochastic local vol hybrid). You need the full surface, not just slices.
• For quick analysis or visualization: Cubic spline is fine as long as you don't trade off it.
• For nothing: Linear interpolation in production. Seriously.

---

Equation Explorer

All interpolation methods work with total variance and log-moneyness. Use this calculator to convert between representations.

Equation Explorer

w = σ² × Ttotal variance = IV² × time

Implied Vol (σ)

%

The implied volatility

Time to Expiry

days

Calendar days to expiration

Total Variance (w)

0.022225

Annualized Variance (σ²)

0.2704

Round-trip IV

52.00%

Total variance is what SVI and other models fit. It scales with time, so a 50% vol for 30 days has less total variance than 50% vol for 90 days.

---

See also:

• How Vol Surfaces Are Built - The full pipeline
• SVI Parameterization - SVI deep dive
• SABR Model - SABR deep dive
• Local Volatility - Dupire deep dive