SVI and Curve Fitting: Building a Modern Implied Volatility Surface

SVI and Curve Fitting in Modern Implied Volatility Modelling

The core challenge in options analytics is constructing a smooth, arbitrage-free implied volatility surface from discrete market quotes. SVI (Stochastic Volatility Inspired) has become widely adopted by professional volatility desks because it captures smile curvature accurately, calibrates quickly, and remains stable during market turbulence.

What SVI Represents

SVI is a parametric formula — not a stochastic volatility model — that maps log-moneyness to total implied variance. The raw SVI parameterization uses five parameters to describe the shape of the volatility smile at a single maturity:

The two key variable definitions underpin the entire framework:

The Five SVI Parameters

Each parameter controls a distinct aspect of the smile shape. Understanding these is essential for diagnosing calibration issues and interpreting the fitted surface:

Why SVI Matters for Options Analytics

Without a smooth, well-calibrated surface, even small bid-ask distortions can create large jumps in Delta, Vega, and fair-value estimates — especially in the wings. SVI eliminates this noise while preserving real market information.

A properly fitted SVI surface enables:

• Consistent Greeks — smooth first and second derivatives of the volatility function yield stable Delta, Gamma, Vega, and Vanna across the entire chain
• Reliable spread valuation — wings, ratio structures, and calendars priced from the same coherent surface
• Risk-neutral density extraction — the second derivative of call prices with respect to strike recovers the implied probability distribution
• Skew and term-structure analytics — quantify skew steepness, convexity, and relative richness across maturities
• Historical surface reconstruction — refit archived option data to study regime changes and volatility dynamics over time

The Calibration Pipeline

Fitting a production-quality SVI surface is not a single optimization step — it is a multi-stage pipeline that must handle noisy data, enforce financial constraints, and scale to thousands of tickers.

1. Filter option quotes

   Discard stale, wide-spread, and illiquid quotes. Use mid-prices where bid-ask is tight; fall back to bid-ask-consistent implied volatility estimates otherwise. Remove obvious outliers that violate basic monotonicity.

2. Convert to total implied variance

   Transform filtered implied volatilities into variance space: $w_i = \sigma^2_{\text{imp},i} \cdot T$. This is the native domain of SVI and linearizes the calendar-spread constraint.

3. Fit SVI parameters per maturity slice

   Run a weighted least-squares optimization for each expiry independently. The objective minimizes the distance between market and model variance:

   Calibration Objective $$\min_{\theta} \sum_{i} w_i \left[w^{\text{market}}_i - w^{\text{SVI}}_i(\theta)\right]^2$$

   where $\theta = (a, b, \rho, m, \sigma)$ and the weights $w_i$ typically reflect quote confidence (tighter spreads get higher weight).

4. Enforce butterfly-arbitrage constraints

   Apply the Gatheral-Jacquier conditions to ensure no negative butterfly spreads exist. This means the risk-neutral density is non-negative everywhere — a hard requirement for any surface used in pricing.

5. Ensure calendar-spread stability

   Verify that total variance is non-decreasing across maturities at every strike: $w(k, T_1) \leq w(k, T_2)$ for $T_1 < T_2$. Violations indicate a calendar-spread arbitrage. Regularize or re-fit adjacent slices when detected.

Ensuring Arbitrage-Free Surfaces

A beautiful fit that violates no-arbitrage conditions is worse than useless — it produces impossible prices and misleading Greeks. The key constraints are:

How SVI Integrates Into a Modern Options Platform

For platforms like FlashAlpha, SVI is not just a fitting exercise — it is the foundation layer that everything else builds on:

• Stable Greeks — smooth derivatives of $w(k)$ translate directly into continuous Delta, Gamma, Vanna, and Volga surfaces
• Accurate spread valuation — consistent skew and wing modelling means complex structures (risk reversals, ratio spreads, butterflies) are priced from a single coherent surface
• Historical surface analytics — refit past data to detect regime shifts, track skew evolution, and identify dislocations
• Quantitative screening — measure skew steepness, convexity shifts, and relative richness across the entire options universe

Limitations and Practical Considerations

• Deep-wing sensitivity — far OTM strikes often have wide spreads and low liquidity; the fitted curve in the wings depends heavily on quote quality and filtering choices
• Extreme-event days — large moves can push the smile into shapes that are hard to capture with five parameters; regularization or parameter bounds may be needed
• Per-maturity independence — each slice is calibrated separately, which can introduce cross-maturity inconsistencies unless calendar constraints are actively enforced
• Model risk — any parametric form imposes structure; SVI cannot capture bimodal implied distributions or discontinuous smile features

Non-Alpha-Sensitive Research Directions

SVI-fitted surfaces open up a rich set of research avenues that do not compromise proprietary trading signals:

• Tracking smile deformation over time — how does the curvature parameter $\sigma$ evolve across market regimes?
• Comparing current skew ($\rho$) to its historical distribution — is today's skew unusually steep or flat?
• Monitoring curvature changes and volatility-of-volatility through parameter time series
• Studying dislocations between traded IV and fitted IV as a measure of market microstructure noise
• Detecting term-structure anomalies — inversions, humps, and kinks that signal event pricing

Conclusion

SVI remains a cornerstone of modern derivatives modelling. Its simplicity, robustness, and arbitrage-free structure make it indispensable for real-time analytics, quantitative research, and options-screening platforms.