SANOS Smooth strictly Arbitrage-free Non-parametric Option Surfaces

Hans Buehler, Blanka Horvath, Anastasis Kratsios, Yannick Limmer, Raeid Saqur

Abstract: We present a simple, numerically efficient but highly flexible non-parametric method to construct representations of option price surfaces which are both smooth and strictly arbitrage-free across time and strike. The method can be viewed as a smooth generalization of the widely-known linear interpolation scheme, and retains the simplicity and transparency of that baseline. Calibration of the model to observed market quotes is formulated as a linear program, allowing bid-ask spreads to be incorporated directly via linear penalties or inequalities, and delivering materially lower computational cost than most of the currently available implied-volatility surface fitting routines. As a further contribution, we derive an equivalent parameterization of the proposed surface in terms of strictly positive "discrete local volatility" variables. This yields, to our knowledge, the first construction of smooth, strictly arbitrage-free option price surfaces while requiring only trivial parameter constraints (positivity). We illustrate the approach using S&P 500 index options.

The Challenge of Option Surface Fitting

In quantitative finance, fitting a continuous surface to discrete market option quotes (bid and ask prices) is a foundational problem. Practitioners require these surfaces to price off-the-run options and extract local volatilities for exotic pricing. However, existing methods force a frustrating trade-off between flexibility, smoothness, and arbitrage guarantees.

  • Parametric models (like SSVI) guarantee smoothness and no-arbitrage but often lack the flexibility to perfectly fit observed market prices, especially in complex market conditions.
  • Non-parametric models (like linear interpolation) can fit the market without arbitrage but produce "kinked" (non-smooth) surfaces. This lack of smoothness breaks continuous-time models like Dupire's local volatility, which rely on second derivatives.
  • Other smooth non-parametric methods often rely on heavy numerical penalties or complex non-linear optimization to suppress arbitrage, making them slow and unreliable for production environments.
Comparison of SANOS with existing interpolation and parametric methods.

The SANOS Methodology

The core technique is a clever generalization of standard linear interpolation. Instead of interpolating prices directly with piecewise linear functions, SANOS represents call option prices as a convex combination of Black-Scholes call payoffs:

$$ \hat{C}(T_j, K) = \sum_{i=0}^N q_j^i \text{Call}(K^i, K, V_j) $$

Here, $K^i$ are quoted strikes, $V_j$ are variances, and $q_j^i$ are weights that act as discrete space transition densities. Because the Black-Scholes pricing function is inherently smooth, this convex combination is also smooth.

To calibrate the model, the authors formulate the search for the weights $q_j$ as a Linear Program (LP). Bid-ask spreads are incorporated directly as simple linear inequalities or penalties, making the optimization globally convex and incredibly fast. Furthermore, the authors introduce a parameterization using Discrete Local Volatilities (DLVs). They prove that to guarantee a strictly arbitrage-free surface, these DLVs only need to be strictly positive—a "trivial" constraint compared to the complex cross-strike and cross-time constraints usually required.

Empirical Performance and Results

The authors validated SANOS using end-of-day S&P 500 (SPX) index option data from 2023 (via OptionMetrics IvyDB). The dataset was filtered to include out-of-the-money (OTM) options with significant liquidity and vega.

SANOS fits for SPX options across 1, 35, and 1249 days to expiry (DTE).

Key results include:

  • Blazing Fast Calibration: Because calibration is a single global linear program, it completes essentially instantaneously, making it viable for high-frequency environments.
  • High-Quality Fit: The model successfully fits SPX market data tightly within the bid-ask spread across a massive range of maturities (from 1 day to 1249 days to expiry).
  • Controllable Smoothness: The model introduces a parameter $\eta$ that controls the smoothness of the fit. Even with high smoothness ($\eta = 0.25$), the maximum fitting error remains well within the bounds of the market bid-ask spread.

Theoretical Insights and Future Work

The DLV parameterization is a significant theoretical contribution for the machine learning community. By ensuring no-arbitrage through simple positivity, SANOS opens the door for neural networks to generate option surfaces easily (e.g., by passing outputs through a Softplus activation function).

While the current framework is designed for European options (indices and futures), future work aims to extend this to American options, which feature early exercise premiums. Additionally, while the standard model is fast, a more "generalized" version allowing different variances per strike currently faces quadratic execution time, presenting an opportunity for further optimization.

Summary updated

Abstract

We present a simple, numerically efficient but highly flexible non-parametric method to construct representations of option price surfaces which are both smooth and strictly arbitrage-free across time and strike. The method can be viewed as a smooth generalization of the widely-known linear interpolation scheme, and retains the simplicity and transparency of that baseline. Calibration of the model to observed market quotes is formulated as a linear program, allowing bid-ask spreads to be incorporated directly via linear penalties or inequalities, and delivering materially lower computational cost than most of the currently available implied-volatility surface fitting routines. As a further contribution, we derive an equivalent parameterization of the proposed surface in terms of strictly positive "discrete local volatility" variables. This yields, to our knowledge, the first construction of smooth, strictly arbitrage-free option price surfaces while requiring only trivial parameter constraints (positivity). We illustrate the approach using S&P 500 index options
BibTeX
@inproceedings{buehler2026sanos,
  author = {Hans Buehler and Blanka Horvath and Anastasis Kratsios and Yannick Limmer and Raeid Saqur},
  title = {{SANOS Smooth strictly Arbitrage-free Non-parametric Option Surfaces}},
  booktitle = {University of Oxford - Symplectic Elements},
  year = {2026},
  url = {https://openreview.net/forum?id=XTZbtCbeb3},
}
Figure