DYSANOS Generative Dynamic Smooth Arbitrage-free Non-parametric Option Surfaces

arXiv:2608.12587 · q-fin.MF, cs.LG · Submitted 2026-08-12 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: AI Radio. Generated commentary on the latest Artificial Intelligence papers.

Tom: We're looking at the paper: "DYSANOS Generative Dynamic Smooth Arbitrage-free Non-parametric Option Surfaces".

Jane: DYSANOS is the first generative market model for smooth SANOS option surfaces for all strikes and expiries which are free of static arbitrage.

Tom: First, who's behind it and why it matters.

Title and authors: Jane: Now, moving on to what they suggest as improvements for DYSANOS, the authors point out that while it’s free of static arbitrage by construction, that doesn't automatically mean it’s free of dynamic arbitrage when options are traded across time periods >

Tom: That’s a crucial distinction. The paper focuses on assessing how static no-arbitrage propagates through those dynamically evolving surfaces using a number of statistical tests defined in Section three point four >

Lu: They compare DYSANOS against a naive AR(one) PCA model applied directly to implied volatilities, and they find that while DYSANOS has no arbitrage when options are held to expiry, evidence of dynamic arbitrage is found when options are traded from one period to the next > <ref:2608.12587#pg1>

Meng: So even though the static part is perfect, the flow of time introduces a problem that needs checking against these specific tests >

Jane: And they also present a conditional spot importance sampling method. This allows them to sample more frequently from the tails of spot movements without changing either the marginal law or any realized path of the surface state >

Tom: That method increases how often spot reaches far strikes, but all the physical-law probabilities and moments use the full mixture weight D, so it keeps things grounded in reality >

Lu: Empirically, they found that this baseline captures first-order level, skew, term, and leverage effects well enough to be useful for initial analysis >

Meng: But they also noted that the Gaussian baseline matches the first moment but misses volatility clustering and crisis tails, so you can't just rely on a simple Gaussian assumption for the whole thing >

Jane: It’s clear they see this as a baseline to beat with more modern time series models, suggesting future work should focus on capturing those secondary directions like volatility clustering >

The paper's summary: Tom: So, wrapping up this section of the talk on DYSANOS Generative Dynamic Smooth Arbitrage-free Non-parametric Option Surfaces, the main point is that they’ve created a model that generates long paths of daily spot and option prices, and every single option surface generated is strictly free of static arbitrage >

Jane: That means you can price any option on any path given those ML-SANOS parameters for that specific sample and time step >

Lu: But the numerical indication they found is that in this situation, DYSANOS does admit dynamic arbitrage along its paths when options are traded from one period to the next >

Meng: So, for deployment purposes, you’d have to be careful because even if the static part is perfect, there can still be exploitable inconsistencies across time steps >

Lalam: I can confirm that DYSANOS is the first generative option market model built from smooth, strictly statically arbitrage-free SANOS surfaces >

Tom: It’s a powerful tool for generating synthetic data, but we need to keep an eye on those dynamic arbitrage tests they ran because that’s where the real complexity lies >

Jane: So, DYSANOS provides a baseline to generate long paths of daily spot and option prices for years in the future, with every option surface free of static arbitrage >

Lu: The approach doesn't rely on approximate static constraints; each smooth option surface is strictly arbitrage free and allows pricing of any option on GPU >

Meng: That GPU part is what makes it interesting for running large-scale simulations without needing a super expensive grid interpolation step >

Lalam: And despite the dynamic arbitrage finding, this paper shows how agents can choose which options to trade at each time step, which is a significant implication for agent-based simulations >

Tom: It sets up the benchmark against pure implied volatility PCA models while showing that DYSANOS exhibits no arbitrage when traded options are held to expiry >

Jane: So, this paper gives us a baseline model to generate long paths of daily spot and option prices, with every option surface free of static arbitrage.

Lu: The approach captures first-order level, skew, term, and leverage effects in the AR(one) structure > <ref:2608.12587#pg1>

Meng: We need to watch how they handle volatility clustering because that’s what’s missing from their baseline compared to more complex dynamics >

Lalam: This paper is DYSANOS Generative Dynamic Smooth Arbitrage-free Non-parametric Option Surfaces.

The paper's improvements: Tom: So we just talked about DYSANOS generating paths for spot prices and option prices that are free of static arbitrage, but now we’re looking at how the authors suggest they can make this thing even better.

Jane: Right. They point out that while the baseline is good, it's missing some of those things we talked about earlier, like volatility clustering and crisis tails.

Tom: Exactly. The baseline captures the first-order effects pretty well, but it understates the skew and leverage effects in those secondary directions.

Lu: Yeah, they suggest that capturing those secondary directions is where a lot of the real interest lies for future work.

Jane: And they show how to do that by looking at things like higher eigensurfaces in the variance of daily changes to spot and hidden state >

Tom: It sounds like they are trying to get better at modeling the volatility itself, not just the average movement.

Lu: They want a model that isn't just smooth; they want one that captures how things change direction over time in a more realistic way.

Jane: And for someone who only listens, what does this mean? It means we need models that can predict when things are going to get wild, not just when the average price will be.

Tom: From an engineering standpoint, they're suggesting ways to tune those dynamics using things like the bandwidth parameter 'w', which lets you mix historical data with continuous noise >

Lu: That way is a good compromise; it keeps the model grounded in what we’ve seen historically while still allowing for some smooth evolution >

Jane: So, it’s not just about making the math prettier; it's about capturing the real-world behavior of markets where things aren't always perfectly smooth.

Tom: And they even show a conditional spot importance sampling method, which lets you sample extreme spot movements more often without messing up the overall path structure >

Lu: That’s smart. If you want to stress test a model, you need to see those tail events more frequently in your simulations >

Jane: So, we're moving from just generating paths to generating paths that are more representative of market stress scenarios.

Tom: And they also push for cost-aware testing where you check if the arbitrage opportunity is actually profitable after accounting for trading costs, like one or ten basis points >

Lu: That makes it way more practical. It moves this from a purely theoretical exercise to something that matters for real trading applications.

Jane: So, the paper isn't just about a new generation model; it’s about building a more realistic and testable framework for synthetic market data.

Tom: It sets up the DYSANOS model as a baseline against pure implied volatility PCA models, which is how you know if this generative approach actually adds value >

Lu: The implication is that the future isn't just about generating surfaces that are arbitrage-free; it’s about generating surfaces that also have realistic dynamics and testable properties.

Conclusion: Tom: So we’re wrapping up our look at DYSANOS Generative Dynamic Smooth Arbitrage-free Non-parametric Option Surfaces, and what we learned is that they built a model that generates long paths of spot prices and option prices where every single surface is strictly free of static arbitrage >

Jane: That means you can price any option on any path given those ML-SANOS parameters for that specific sample and time step, which is a huge deal for simulation work >

Lu: Yeah, the authors argue this isn't just about creating pretty surfaces; it’s about building something where every option has a unique price on every path >

Meng: From what I see, the main thing to watch is that while the static arbitrage is gone, they still found evidence of dynamic arbitrage when you trade from one time period to the next >

Tom: Exactly. So, even with that static guarantee, there’s still an issue when options are traded over time >

Jane: And they showed conditional spot importance sampling as a way to sample extreme movements more often without breaking the overall path structure >

Lu: That method is clever because it lets you focus on the tails—the rare events—without altering the fundamental laws of probability for those moments >

Meng: From a production standpoint, I wonder how much computational overhead that sampling adds when we’re trying to generate millions of paths at scale >

Tom: That’s fair. It adds complexity to the generation process, and we need to see if the efficiency gain from better tail coverage outweighs that added runtime >

Jane: So, what does this mean for us listeners who aren't deep into the math? It means we have a tool that can create synthetic market data where every single option is priced correctly on any given day >

Lu: It’s about moving away from models that assume everything is perfectly smooth and instead embracing dynamics that capture how things actually move in the real world >

Meng: I think the big picture here is using these generative models to stress test trading strategies under realistic, complex conditions, not just perfect theoretical ones >

Tom: So this paper, DYSANOS Generative Dynamic Smooth Arbitrage-free Non-parametric Option Surfaces, gives us a starting point for generating those long paths where every option surface is free of static arbitrage >

Jane: It’s a great tool for stress testing and simulation work, and while it admits dynamic arbitrage along its paths, the framework itself is very solid >

Lu: The authors flag that capturing volatility clustering and crisis tails are the next big targets for improving this baseline model >

Meng: I think if we can integrate those dynamic elements smoothly, this could become a really powerful engine for risk management simulations >

University of Oxford · University of Oxford · McMaster University

q-fin.MF, cs.LG

Submitted: 2026-08-12

Updated: 2026-10-15

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 75/100

The gist: DYSANOS is the first generative market model for smooth SANOS option surfaces for all strikes and expiries which are free of static arbitrage.

Key concepts

Static Arbitrage: A set of option prices is free of static arbitrage if no portfolio of options held to expiry can yield a riskless profit. This requires prices to be consistent across strikes and expiries, e.g., monotonicity and convexity in strike. DYSANOS guarantees this for every generated surface.
Dynamic Arbitrage: Dynamic arbitrage occurs when trading options across multiple periods can yield a riskless profit, even if each period's surface is statically arbitrage-free. It arises from inconsistencies between consecutive surfaces. DYSANOS exhibits dynamic arbitrage when options are traded from one period to the next.
SANOS Model: SANOS is a parametric representation of smooth option surfaces that is free of static arbitrage by construction. It maps a low-dimensional state to call prices for all strikes and expiries. DYSANOS uses a neural decoder (ML-SANOS) to transform a latent state into a SANOS surface.
AR(1) Factor Model: An autoregressive model of order one for the latent state, with mean reversion and PCA-driven noise. It captures daily changes in spot and surface. The bandwidth parameter mixes historical and continuous noise. This baseline reproduces first-order level, skew, term, and leverage effects but understates them.

Terminology

Clustering — 4× in this paper · explained in 2 episodes
Generative Model — 4× in this paper · explained in 2 episodes
alpha — 4× in this paper
Calibration — 3× in this paper · explained in 9 episodes
Benchmark — 2× in this paper · explained in 2 episodes
Activation Function — 1× in this paper · explained in 1 episode
Actor — 1× in this paper · explained in 3 episodes
Alignment — 1× in this paper
Consistency — 1× in this paper · explained in 3 episodes
Generalization — 1× in this paper · explained in 13 episodes
Generative Models — 1× in this paper · explained in 3 episodes
Wasserstein Distance — 1× in this paper · explained in 5 episodes

Summary

DYSANOS is the first generative market model for smooth SANOS option surfaces for all strikes and expiries which are free of static arbitrage. The model is designed to generate entire paths of daily spot and option prices for years in the future. The paper presents a robust and useful if somewhat simplistic baseline in the form of an AR(1) model, discusses model setup, data pipeline, and training, and investigates numerical presence of dynamic arbitrage. Model performance is illustrated on Option Metrics' IvyDB S&P Index data from 2020 to 2025 and compared to a pure implied-vol PCA model.

The paper introduces a ML-SANOS decoder, a coordinate transformation of the SANOS model such that the state parameters ht live in a low-dimensional space Rnh for nh 20. The call prices remain smooth and statically arbitrage free. This ML-SANOS is a natural decoder of a driving state process into an option surface of the form ht → C(ht; T, K). It is shown how ML-SANOS can be trained directly to suitably pre-processed option market data, yielding a trained ht for each day. The pre-processing step means that DYSANOS, via SANOS, is fitted to real market bid/ask prices.

For generative dynamics, the paper shows that most of the variance of daily changes to the joint space of spot and hidden state can be explained by essentially five parameters, in line with general PCA analysis of the implied vol surface. A simple baseline version of a generative state space model is presented based on a simple auto-regressive low-dimensional factor model which has in spirit the form dh̃t+1 = κ(m − h̃t)dt + wαt′(ω) + 1 − w2 AYt(ω) dut dt, where m ∈ Rnh is the long-term mean, κ ∈ Rnh ×nh is the mean-reversion speed and dut ∈ Rnα,nh are the main PCA factors driving dh̃t. The matrix A is the Cholesky decomposition of the historic covariance of the factor loadings α ∈ Rnt,nα. Moreover, Yt(ω) ∈ Rnα is normal and αt(ω) ∈ Rnα is sampled from historic factor loadings (αt)t∈T. The bandwidth parameter w can be used to mix between purely historical sampling with w = 1 and fully continuous state sampling w = 0. The actual model presented has an additional state h̃0 without mean-reversion which represents log-spot; it is also simulated using a more robust representation for large time steps. The authors do not claim that this approach provides the most realistic dynamics; rather, they see this implementation as baseline to beat with more modern time series models.

Regarding absence of dynamic arbitrage, every decoded SANOS surface is free of static arbitrage by construction. However, this cross-sectional property does not by itself imply the absence of dynamic arbitrage. Encoding dynamic no-arbitrage is particularly challenging for generated discrete data, since conditional means are typically estimated from noisy samples, making violations difficult to assess reliably. The paper focuses on assessing how static no-arbitrage propagates through the dynamically evolving call surfaces. A number of statistical tests are defined in Section 3.4 to detect the presence of actionable arbitrage in the generated data. DYSANOS is compared with a naive AR(1) PCA model applied directly to implied volatilities. While DYSANOS, as expected, exhibits no arbitrage when traded options are held to expiry, evidence of dynamic arbitrage is nevertheless found when options are traded from one period to the next.

The paper also presents a conditional spot importance sampling method. The surface-first structure allows sampling more frequently from the tails of spot without changing either the marginal law or any realized path of the surface state. The method increases the frequency with which spot reaches far strikes, but all reported physical-law probabilities and moments use the full mixture weight D. No marginal correction is applied in isolation, and the surface path is never resampled or altered by the proposal.

Empirical diagnostics show that the baseline captures the direction and most of the magnitude of the first-order level, skew, term, and leverage effects, but understates them. The baseline reproduces the dominant surface-change direction closely, but concentrates too much variance in it and is less accurate for secondary directions. The Gaussian baseline matches the first moment but not volatility clustering and crisis tails.

In testing for arbitrage in generated data, for returns to expiry, DYSANOS produces no candidates in the test, as expected. The PCA-IV benchmark contains complete-support arbitrage at zero cost and after one- and ten-basis-point costs. For period returns, the larger universe resolves every selected state–time center. DYSANOS has the higher occurrence in every row, and both generators retain candidates at ten basis points. An exact-state conditional resampling of the two strongest DYSANOS weekly raw period-return candidates at ten basis points rejects one of the two strongest DYSANOS candidates and leaves the other unresolved at the stated sampling precision. It does not establish non-negative P&L on the complete conditional support, but, considering the evidence, is a strong indicator that DYSANOS does admit dynamic arbitrage.

The conclusion states that DYSANOS is the first generative option market model built from smooth, strictly statically arbitrage-free SANOS surfaces. The approach does not rely on approximate static constraints; each smooth option surface is strictly arbitrage free and allows pricing of any option on GPU. The surface-first AR(1) captures first-order level, skew, term, and leverage effects, while the empirical diagnostics identify volatility clustering, crisis tails, and higher eigensurfaces as natural targets for future work. The paper provides a baseline model to generate long paths of daily spot and option prices. Each option surface is free of static arbitrage. Moreover, any option can be priced on each path given the ML-SANOS parameters for the respective sample and time step. Contrary to previous results, agents can therefore choose which options to trade at each time step. Numerical indication is found that in this situation, DYSANOS admits dynamic arbitrage along its paths.

Improvements for AI systems

Improvements to AI Systems Based on This Paper:

Arbitrage-Free Generative Market Modeling

Implement a neural decoder (ML-SANOS) that maps low-dimensional latent states to smooth, strictly static-arbitrage-free option surfaces across all strikes and expiries.
The improved AI system can generate entire multi-year paths of daily spot prices and full option surfaces without violating cross-sectional no-arbitrage constraints, enabling realistic synthetic market data for stress testing, backtesting, and simulation.

Dynamic Arbitrage Detection and Quantification

Integrate the paper’s statistical tests (Section 3.4) for detecting actionable dynamic arbitrage in generated data, including cost-adjusted and conditional-support checks.
The improved system can automatically flag whether a generative model admits dynamic arbitrage when options are traded across periods, allowing users to reject or adjust models that produce exploitable inconsistencies.

Low-Dimensional State-Space Representation with Mean Reversion

Use the AR(1)-style factor model with mean reversion (κ, m) and PCA-driven noise (A, α, Y) to capture first-order level, skew, term, and leverage effects in option surfaces.
The improved system can generate realistic daily changes in spot and surface states with reduced dimensionality (20 latent variables), making it computationally efficient for large-scale Monte Carlo simulations on GPUs.

Conditional Importance Sampling for Tail Events

Adopt the surface-first conditional spot importance sampling method to oversample extreme spot movements without altering marginal laws or surface paths.
The improved system can efficiently estimate rare-event probabilities (e.g., deep out-of-the-money option payoffs, crash scenarios) with higher precision, while preserving unbiased physical-law moments via mixture weights.

Benchmarking and Diagnostics for Generative Models

Use the paper’s empirical diagnostics (first-order effects, variance concentration, volatility clustering, crisis tails) to evaluate and compare generative models.
The improved system can automatically diagnose model weaknesses—such as understated skew or over-concentrated variance—and guide iterative improvements toward more realistic dynamics.

Cost-Aware Arbitrage Testing in Generated Data

Incorporate transaction cost thresholds (e.g., 1 and 10 basis points) into arbitrage detection, as done in the paper.
The improved system can determine whether apparent arbitrage opportunities remain profitable after realistic trading costs, providing a practical measure of model realism for trading applications.

Pathwise Option Pricing and Trading Simulation

Leverage the ML-SANOS decoder to price any option (arbitrary strike/expiry) at each time step along generated paths.
The improved system enables agent-based simulations where trading decisions (which options to buy/sell) are made dynamically at each step, supporting realistic portfolio optimization and risk management research.

Hybrid Historical-Continuous Noise Sampling

Implement the bandwidth parameter w to mix historical factor loadings with continuous Gaussian noise, balancing realism and smoothness in generated dynamics.
The improved system can tune w to match desired volatility clustering or crisis behavior, offering flexibility for different simulation purposes (e.g., stress testing vs. steady-state generation).

Model Comparison and Selection Framework

Use the paper’s comparison against a pure implied-vol PCA model to establish a baseline for evaluating new generative approaches.
The improved system can automatically benchmark novel time-series models (e.g., transformers, normalizing flows) against this baseline, quantifying improvements in no-arbitrage preservation, moment matching, and arbitrage occurrence rates.

GPU-Accelerated Synthetic Data Generation

Exploit the smooth, parametric SANOS representation to price options on GPU without grid interpolation.
The improved system can generate millions of realistic market paths in parallel, enabling large-scale reinforcement learning, option pricing, and risk analytics that require massive synthetic datasets.

Abstract

This article presents with DYSANOS the first generative market model for smooth SANOS option surfaces for all strikes and expiries which are free of static arbitrage. Our model is designed to generate entire paths of daily spot and option prices for years in the future. We present a robust and useful if somewhat simplistic baseline hidden state generative model in the form of an AR(1) model. We discuss model setup, data pipeline, and training and investigate numerical resence of dynamic arbitrage. We illustrate model performance on Option Metrics' IvyDB S&P Index data from 2020 to 2025 and compare it to a pure implied-vol PCA model.