Stochastic Volatility Inspired (SVI)

The foundational formulation expresses total implied variance as a function of log-moneyness (k). It is defined by five intuitive parameters {a, b, ρ, m, σ}.

• a: Controls the overall level of variance.
• b: Defines the angle between the left and right asymptotes.
• ρ: The correlation parameter, which induces the skew or asymmetry of the smile.
• m: Shifts the smile horizontally along the log-moneyness axis.
• σ: Governs the smoothness or curvature of the smile's minimum point.

The raw form ensures the curve is strictly convex and behaves linearly in the tails, preventing static arbitrage.

Unlike complex stochastic volatility models that require computationally expensive numerical methods, SVI offers a quasi-explicit calibration process. The parameters can be estimated using a two-step procedure:

• Step 1: Fit the raw SVI parameters to observed market data for a single maturity using a non-linear least-squares optimizer.
• Step 2: Apply explicit transformations to convert raw parameters into the 'natural' SVI parameters, which have clearer geometric interpretations. This efficiency makes it ideal for high-frequency trading systems where speed is critical.

A critical advantage of the SVI model is its ability to guarantee the absence of static arbitrage (butterfly and calendar spread arbitrage) through explicit parameter constraints.

• No Butterfly Arbitrage: Ensured by verifying that the implied density is non-negative. This translates to a simple condition on the minimum of the function g(k) derived from the SVI parameters.
• No Calendar Spread Arbitrage: Guaranteed by ensuring total implied variance is a non-decreasing function of time to maturity. This is managed by fitting slices independently and then interpolating total variance linearly. These conditions are mathematically tractable, unlike in many other parametric forms.

While SVI is defined for a single maturity slice, a full volatility surface is constructed by interpolating between independently calibrated slices.

• Total Variance Interpolation: To prevent calendar spread arbitrage, the interpolation is performed on total implied variance (w = σ² * T) rather than on implied volatility directly.
• Slice-by-Slice Calibration: Each maturity's SVI parameters are calibrated independently, allowing the model to capture distinct market regimes across different tenors.
• Smoothing: The resulting surface is smooth in the strike direction by construction and requires only a monotonic interpolation scheme in the time direction.

The 'Inspired' in SVI refers to its deep mathematical connection to the Heston stochastic volatility model. For large maturities, the implied volatility smile generated by the Heston model converges exactly to an SVI-type parameterization.

• The SVI parameters map directly to Heston's parameters: correlation (ρ), mean-reversion speed, and volatility-of-volatility.
• This provides a microstructural foundation for the parameterization, linking it to the dynamics of the underlying spot and variance processes.
• It allows traders to use SVI as a fast, robust proxy for the full Heston model in pricing and risk management.

To better capture the extreme curvature observed in short-dated options, the Jump-Wings (JW-SVI) extension adds additional parameters to model the behavior of very deep out-of-the-money options.

• It explicitly parameterizes the left and right tail asymptotes, allowing for more aggressive or dampened wing slopes than the linear tails of the raw SVI.
• This is crucial for accurately pricing tail risk and deep out-of-the-money puts, which are sensitive to crash risk.
• The extension maintains the arbitrage-free properties while providing a superior fit for short-maturity smiles where jumps in the underlying are a dominant risk factor.

The graphical phenomenon where out-of-the-money and in-the-money options exhibit higher implied volatilities than at-the-money options for a fixed expiration. SVI directly parameterizes this curve.

• Violates the constant volatility assumption of the Black-Scholes model
• Reflects market expectations of fat tails and crash risk
• The smile's shape varies by asset class: equity smiles are typically downward-sloping (skew), while FX smiles are symmetric

Mathematical conditions ensuring that an implied volatility parameterization does not permit risk-free profits through calendar or butterfly spreads. A core advantage of SVI is its ability to enforce these constraints analytically.

• Calendar arbitrage: Total implied variance must be monotonically increasing with time to expiration
• Butterfly arbitrage: The risk-neutral probability density function must remain non-negative
• SVI's parametric form allows for direct constraint on the Gatheral conditions

The raw parameterization of total implied variance w(k) as a function of log-moneyness k, defined by Jim Gatheral in 2004. It uses five intuitive parameters to capture the smile's shape.

• a: Controls the overall level of variance
• b: Governs the angle between the left and right asymptotes
• ρ: Defines the skew or rotational tilt of the smile
• m: Shifts the smile horizontally along the moneyness axis
• σ: Controls the smoothness or curvature at the at-the-money point

A three-dimensional representation of implied volatility across both strike price and time to expiration. SVI is typically calibrated independently for each maturity slice, which are then interpolated to form the full surface.

• Axes: Moneyness (x), Time to Expiry (y), Implied Volatility (z)
• Used to price exotic options and calculate local volatility via Dupire's formula
• A smooth, arbitrage-free surface is the primary deliverable of an SVI calibration engine

A fast calibration technique for the SVI Jump-Wings (SVI-JW) parameterization that reduces the optimization problem to a series of linear regressions rather than a slow, non-convex global search.

• Avoids the pitfalls of local minima common in brute-force optimization
• Enables real-time recalibration for market-making systems
• The method exploits the geometric interpretation of SVI parameters to fit the observed smile in milliseconds

A deterministic function σ(S,t) that extends the SVI concept by making volatility a function of both the underlying asset price and time. The SVI-parameterized implied volatility surface is the critical input for deriving this model via Dupire's formula.

• Ensures exact calibration to the current market smile
• Used for pricing path-dependent options like barriers and cliquets
• The local volatility surface is numerically computed from the first and second derivatives of the SVI implied variance