Regime-based Implied Stochastic Volatility Model for Crypto Option Pricing

Danial Saef, Yuanrong Wang, Tomaso Aste

TL;DR

This paper introduces a novel approach to pricing cryptocurrency options by integrating market regime clustering with implied stochastic volatility models. By segmenting the highly volatile crypto market into distinct volatility regimes, the authors achieve more accurate option pricing, overcoming limitations of traditional models that struggle with non-stationarity and jumps.

What Problem Does This Paper Address?

The burgeoning cryptocurrency market, particularly Bitcoin, presents a significant challenge for financial modeling due to its extreme volatility, frequent price jumps, and non-stationary nature. Traditional option pricing models, often fitted to historical return data, fail to capture these dynamics, leading to inaccurate pricing. Existing advanced models, like Implied Stochastic Volatility Models (ISVMs), can incorporate investor expectations via implied volatility (IV) but struggle to handle the complexities of crypto market microstructure, especially jumps. Furthermore, the rapid evolution of crypto markets means historical patterns may not reliably predict future behavior, making purely data-driven approaches problematic.

Methods

The core innovation is the Regime-based Implied Stochastic Volatility Model (MR-ISVM). This approach combines two key techniques:

  • Inverse Covariance Clustering (ICC): This is a temporal clustering algorithm adapted for this study. Instead of relying on historical price data, ICC uses the implied volatility (IV) surface (specifically, characteristics like moneyness and time-to-maturity) to identify distinct market regimes. It penalizes frequent switches between clusters, promoting temporal consistency. This method helps to segment the data into periods with stable volatility characteristics without needing to explicitly model complex jump dynamics.
  • Implied Stochastic Volatility Modeling (ISVM): Building on existing ISVM frameworks, the authors fit separate ISVMs to each identified market regime. This allows the model to adapt to the specific volatility dynamics of each regime, leveraging investor expectations embedded in the IV surface. The ISVM framework itself is a continuous bi-variate model that can, in theory, account for jumps, but the paper focuses on using clustering to implicitly handle non-stationarity and jump-like behavior.

Datasets

The study utilizes high-frequency cryptocurrency option data from Deribit, a popular trading platform. Specifically, they use data from April 13, 2021, to March 7, 2022, with a sampling frequency of 20 minutes. This dataset comprises over 9 million observations across 12,000+ assets.

Key Insights and Ideas

  • Investor Expectations as a Key Driver: The paper emphasizes that in the volatile crypto market, investor sentiment and expectations, as reflected in implied volatility, are more crucial for pricing than historical returns alone.
  • Clustering as a Proxy for Jumps: By clustering the IV surface into distinct regimes, the MR-ISVM implicitly accounts for the non-stationarity and jump-like behavior characteristic of crypto markets, without the empirical challenges of directly modeling jumps.
  • Adaptability is Crucial: The dynamic nature of crypto markets necessitates models that can adapt to changing conditions. The regime-based approach provides this adaptability.
  • Short-Maturity Options Reflect Current Sentiment: The study leverages short-maturity options (one week or less) for clustering, hypothesizing that these best capture immediate investor beliefs and market sentiment.

Key Results

  • Reduced Estimation Error: The MR-ISVM approach, by using clustering, significantly reduces the Root Mean Squared Error (RMSE) for estimating the model's parameters compared to fitting a single model on the entire dataset.
  • Improved Stability: The clustering approach also leads to a more stable estimation of model parameters, as indicated by a narrower spread in RMSE values across different clusters.
  • Distinct Regimes Identified: The analysis of Bitcoin options revealed distinct volatility regimes. For instance, with two clusters (K=2), one regime captured periods of higher volatility and price movements, while the other represented more stable periods.
  • Model Fit Improvement: Visualizations show that the MR-ISVM provides a better fit to the underlying functions compared to the unclustered model, with distinct patterns captured for each regime.

Limitations and Future Work

  • Choice of Parameters: The performance of the MR-ISVM is sensitive to the choice of clustering parameters, specifically K (the number of clusters) and λ (the temporal consistency penalty). The paper explored these but suggests further systematic examination.
  • Benchmarking: While the paper demonstrates the effectiveness of MR-ISVM, it acknowledges the need for benchmarking against other commonly used models in the literature and industry, such as Stochastic Volatility with Correlated Jumps (SVCJ).
  • Interpretability of Clusters: While the paper explores the effect of clustering on model interpretability, interpreting the exact meaning of clusters can still be challenging, especially with a higher number of clusters.
  • Future Research: The authors suggest exploring the optimal choice of K and λ, and comparing MR-ISVM with models like SVCJ to further validate its performance and understand its advantages in the context of crypto option pricing.

Summary updated

Abstract

The increasing adoption of Digital Assets (DAs), such as Bitcoin (BTC), rises the need for accurate option pricing models. Yet, existing methodologies fail to cope with the volatile nature of the emerging DAs. Many models have been proposed to address the unorthodox market dynamics and frequent disruptions in the microstructure caused by the non-stationarity, and peculiar statistics, in DA markets. However, they are either prone to the curse of dimensionality, as additional complexity is required to employ traditional theories, or they overfit historical patterns that may never repeat. Instead, we leverage recent advances in market regime (MR) clustering with the Implied Stochastic Volatility Model (ISVM). Time-regime clustering is a temporal clustering method, that clusters the historic evolution of a market into different volatility periods accounting for non-stationarity. ISVM can incorporate investor expectations in each of the sentiment-driven periods by using implied volatility (IV) data. In this paper, we applied this integrated time-regime clustering and ISVM method (termed MR-ISVM) to high-frequency data on BTC options at the popular trading platform Deribit. We demonstrate that MR-ISVM contributes to overcome the burden of complex adaption to jumps in higher order characteristics of option pricing models. This allows us to price the market based on the expectations of its participants in an adaptive fashion.
BibTeX
@inproceedings{saef2023regime,
  author = {Danial Saef and Yuanrong Wang and Tomaso Aste},
  title = {{Regime-based Implied Stochastic Volatility Model for Crypto Option Pricing}},
  booktitle = {The Web Conference},
  year = {2023},
  doi = {10.1145/3543873.3587621},
}
Figure