Paper: arXiv 2311.04727

Authors: Siu Hin Tang, Mathieu Rosenbaum, Chao Zhou

Abstract

We extend the application and test the performance of a recently introduced volatility prediction framework encompassing LSTM and rough volatility. Our asset class of interest is cryptocurrencies, at the beginning of the “crypto-winter” in 2022. We first show that to forecast volatility, a universal LSTM approach trained on a pool of assets outperforms traditional models. We then consider a parsimonious parametric model based on rough volatility and Zumbach effect. We obtain similar prediction performances with only five parameters whose values are non-asset-dependent. Our findings provide further evidence on the universality of the mechanisms underlying the volatility formation process.

Complexity vs Empirical Score

  • Math Complexity: 8.0/10
  • Empirical Rigor: 6.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced mathematical concepts like quadratic rough Heston, fractional Brownian motion, and the Zumbach effect, while also conducting a comprehensive empirical analysis on a large dataset of 213 cryptocurrencies with specific train-test splits and rigorous evaluation metrics (MSE).

Research Flowchart

  flowchart TD
  A["Research Goal<br>Forecast Cryptocurrency Volatility<br>using ML & Rough Volatility"] --> B{"Methodology"};
  B --> C["Universal LSTM Model<br>Trained on Asset Pool"];
  B --> D["Rough Volatility &<br>Zumbach Effect Model"];
  C --> E["Data: BTC/ETH/Altcoins<br>2022 Crypto-Winter Period"];
  D --> E;
  E --> F["Computational Process"];
  F --> G["Comparison of<br>Prediction Performance"];
  G --> H["Key Findings/Outcomes<br>Universal LSTM Outperforms Traditional Models<br>Rough Volatility Model Achieves Similar Performance<br>with Only 5 Non-Asset-Dependent Parameters"];

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a universal LSTM volatility forecaster trained on pooled cryptocurrency data and a parsimonious rough‑volatility parametric model with a Zumbach effect.

Source sentence
We first show that to forecast volatility, a universal LSTM approach trained on a pool of assets outperforms traditional models. We then consider a parsimonious parametric model based on rough volatility and Zumbach effect.

Assumes 6

  • Cryptocurrencies trade every day, so HAR windows are set to 7 and 30 days. market
    Source sentence
    Due to the fact that cryptocurrencies are traded every day, for the HAR, the windows are chosen to be 7 and 30 instead of 5 and 22 in the literature on traditional markets.
  • Only coins that are not stablecoins, leveraged, low‑liquidity or have insufficient data are kept. data
    Source sentence
    We select 213 coins after filtering out coins that are stablecoins, leveraged, or have low liquidity or insufficient data.
  • Training data start from 2020; earlier data are excluded. data
    Source sentence
    We choose not to use training data from before 2020.
  • Training period is 2020‑01‑01 to 2021‑12‑31; test period is 2022‑01‑01 to 2022‑06‑30. data
    Source sentence
    The training set is chosen to be the two-year period from 2020-01-01 to 2021-12-31, and the test set is a six-month period from 2022-01-01 to 2022-06-30.
Show 2 more
  • Volatility forecasts are evaluated with mean‑squared‑error (MSE). statistical
    Source sentence
    For model evaluation, we choose to use the mean-squared-error or MSE, calculated as MSE(σ, ˆσ) = 1/T
  • Volatilities and returns are scaled and normalized by each coin’s training‑set statistics. data
    Source sentence
    The volatilities and returns are scaled and normalized respectively by the individual coin’s statistics calculated on the training set.

Data

  • Binance 5‑minute intraday price data · Binance.com market-prices public
    2020‑01‑01 to 2022‑06‑30 · 213 selected cryptocurrencies (excluding stablecoins, leveraged, low‑liquidity)
    Source sentence
    Our dataset contains 5-minute intraday prices of cryptocurrencies traded on the Binance exchange, with USDT, a popularly used stablecoin, as the base currency.

Recognised sources: Binance

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • Universal LSTM models outperform the HAR baseline for most coins. performance
    Source sentence
    LSTMs outperform HAR in general.
  • The rough‑volatility RFSV model outperforms HAR and is comparable to LSTM variants. performance
    Source sentence
    RFSV outperforms HAR, is slightly better than LSTM7,var and is close to LSTM30,var, but underperforms LSTM7,ret and LSTM30,ret.
  • Adding past returns via the QRH component improves RFSV, matching LSTM30,ret at λ≈0.15. performance
    Source sentence
    Incorporating past returns in the RFSV through the QRH component is a clear improvement. At λ = 0.15, the model performs best and is very similar to LSTM30,ret.
  • Both non‑parametric LSTM and parsimonious RFSV outperform all traditional industry models. performance
    Source sentence
    The non‑parametric LSTM and parsimonious RFSV, like for US stocks, outperform all other traditional models used in industry and past literature.
  • The volatility formation process appears universal across a wide selection of cryptocurrencies, similar to stocks. theoretical
    Source sentence
    We see that LSTMs and RFSV, which are calibrated with the pooled dataset, outperform linear models calibrated specifically to each coin. This not only supports the use of the two non‑linear model types, but also the suspected universality in volatility which is captured by these models.

Checked by others

Supported by

  • To be or not to be: Roughness or long memory in volatility? (2024)
    On the question: Do cryptocurrencies exhibit a universal, rough volatility formation process?
    Compare the findings

    This paper: The volatility formation process appears universal across a wide selection of cryptocurrencies, similar to stocks.

    That paper: The log‑variance of large‑cap cryptocurrency spot rates exhibits roughness, aligning with prior findings on volatility formation.

    Both findings report rough volatility in crypto; B confirms A’s universal formation claim for large‑cap spot rates, showing consistent roughness across samples.

Cited by 1 paper in the archive

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (44)

  1. On the universality of the volatility formation process: when machine learning and rough volatility agree (2022)
  2. Deep calibration of the quadratic rough H eston model (2022)
  3. A simple approximate long-memory model of realized volatility (2009)
  4. Bitcoin: A peer-to-peer electronic cash system (2008)
  5. Long short-term memory (1997)
  6. Adam: a method for stochastic optimization (2015)
  7. Gaussian Error Linear Units (GELUs) (2020)
  8. Volatility is rough (2018)
and 36 more
  1. Cryptocurrency trading: a comprehensive survey (2022)
  2. Volatility forecast comparison using imperfect volatility proxies (2011)
  3. What is stablecoin?: A survey on price stabilization mechanisms for decentralized payment systems (2019)
  4. A hybrid volatility forecasting framework integrating GARCH , artificial neural network, technical analysis and principal components analysis (2018)
  5. Asymmetric volatility dynamics in cryptocurrency markets on multi-time scales (2022)
  6. Hybrid deep learning and GARCH -family models for forecasting volatility of cryptocurrencies (2023)
  7. Deep learning in predicting cryptocurrency volatility (2022)
  8. A GMM approach to estimate the roughness of stochastic volatility (2023)
  9. Factors affecting institutional investors to add crypto-currency to asset portfolios (2021)
  10. On the performance of cryptocurrency funds (2022)
  11. Investigating e-retailers’ intentions to adopt cryptocurrency considering the mediation of technostress and technology involvement (2022)
  12. Rough volatility of B itcoin (2020)
  13. Asymmetric volatility in cryptocurrency markets: New evidence from smooth transition GARCH models (2020)
  14. Asymmetric volatility in cryptocurrencies (2018)
  15. Pricing under rough volatility (2016)
  16. Roughening H eston (2019)
  17. The characteristic function of rough H eston models (2019)
  18. Perfect hedging in rough H eston models (2018)
  19. Deep learning (2016)
  20. Decoupling the short-and long-term behavior of stochastic volatility (2022)
  21. Roughness in spot variance? A GMM approach for estimation of fractional log-normal stochastic volatility models using realized measures (2020)
  22. A million metaorder analysis of market impact on the B itcoin (2015)
  23. Bitcoin: A natural oligopoly (2022)
  24. Why B itcoin will fail to scale? (2022)
  25. Is B itcoin really untethered? (2020)
  26. From rough to multifractal volatility: The log S -f BM model (2022)
  27. Quantifying reflexivity in financial markets: Toward a prediction of flash crashes (2012)
  28. Apparent criticality and calibration issues in the Hawkes self-excited point process model: application to high-frequency financial data (2015)
  29. Cryptocurrency pump-and-dump schemes (2021)
  30. The psychology of cryptocurrency trading: Risk and protective factors (2021)
  31. Quadratic Hawkes processes for financial prices (2017)
  32. From quadratic Hawkes processes to super- H eston rough volatility models with Zumbach effect (2021)
  33. The microstructural foundations of leverage effect and rough volatility (2018)
  34. No-arbitrage implies power-law market impact and rough volatility (2020)
  35. Volatility conditional on price trends (2010)
  36. The quadratic rough H eston model and the joint S & P 500/ VIX smile calibration problem (2020)

0 of 44 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.