Paper: arXiv 2309.06393
Authors: Yutong Chen, Paul Bilokon, Conan Hales, Laura Kerr
Abstract
Cryptocurrency market is known for exhibiting significantly higher volatility than traditional asset classes. Efficient and adequate risk calculation is vital for managing risk exposures in such market environments where extreme price fluctuations occur in short timeframes. The objective of this thesis is to build a real-time computation workflow that provides VaR estimates for non-linear portfolios of cryptocurrency derivatives. Many researchers have examined the predictive capabilities of time-series models within the context of cryptocurrencies. In this work, we applied three commonly used models - EMWA, GARCH and HAR - to capture and forecast volatility dynamics, in conjunction with delta-gamma-theta approach and Cornish-Fisher expansion to crypto derivatives, examining their performance from the perspectives of calculation efficiency and accuracy. We present a calculation workflow which harnesses the information embedded in high-frequency market data and the computation simplicity inherent in analytical estimation procedures. This workflow yields reasonably robust VaR estimates with calculation latencies on the order of milliseconds.
Complexity vs Empirical Score
- Math Complexity: 6.0/10
- Empirical Rigor: 7.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper uses advanced volatility models (GARCH, HAR) and expansions (Cornish-Fisher) but grounds them in a pragmatic, implementation-focused workflow; the significant empirical rigor comes from the explicit kdb+/q implementation, discussion of latency (14.2ms), and benchmarking against real crypto market data.
Research Flowchart
flowchart TD A["Research Goal:<br>Real-time VaR for<br>Non-linear Crypto Derivatives"] --> B["Data Input:<br>High-Frequency<br>Crypto Market Data"] B --> C["Methodology:<br>Volatility Models<br>EWMA, GARCH, HAR"] C --> D["Computational Process:<br>Delta-Gamma-Theta<br>plus Cornish-Fisher Expansion"] D --> E["Outcomes:<br>Robust VaR Estimates<br>in Milliseconds"]
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 real-time VaR calculation workflow for cryptocurrency derivative portfolios using kdb+ and analytical approximations.
Source sentence
The objective of this thesis is to build a real-time computation workflow that provides VaR estimates for non-linear portfolios of cryptocurrency derivatives.
Assumes 6
- A frictionless market is assumed for realised variance to be a consistent estimator of quadratic variance. market
Source sentence
For realised variance to be used as a consistent estimator of quadratic variance (QV), an important assumption is a frictionless market, or in stochastic process term, the log-price process must be a continuous semimartingale.
- Error terms in GARCH models are assumed to follow non‑Gaussian distributions such as Student‑t. model
Source sentence
it is common to assume error terms follow non-Gaussian distributions, such as Student-t, skewed Student-t and generalised normal distribution.
- High‑frequency market data are assumed to be available and can be processed in real time. data
Source sentence
We present a calculation workflow which harnesses the information embedded in high-frequency market data and the computation simplicity inherent in analytical estimation procedures.
- Delta‑gamma‑theta approach approximates the tail behaviour of the portfolio return distribution. model
Source sentence
To further reduce the calculation latency, delta-gamma-theta approach is chosen as a replacement of the commonly used historical or Monte Carlo simulation approach, to approximate the non-Gaussian distribution of portfolio returns.
Show 2 more
- The covariance matrix elements are forecasted independently, implying positive semi‑definiteness is not guaranteed. model
Source sentence
In the context of our inference algorithm, the elements within the covariance matrix are forecasted independently.
- Parallel processing is assumed to be available but only minimally exploited due to the limited number of underlyings. execution
Source sentence
kdb+ is capable of supporting parallel processings with threads and multi-processes. In this work, we could only exploit this benefit to a minimal extent, given there are only two indices within the scope of this work.
Data
- Deribit exchange data options unknown
Source sentence
In this thesis, financial time series data are sourced from Deribit exchange, a leading cryptocurrency exchange that specialises in the futures and European-style options market.
Costs
Not applicable
Finds 5
- The HAR model achieved the best inference accuracy, passing all back‑test criteria (9/9). performance
Source sentence
Overall, HAR has the best performance in terms of accuracy.
- EWMA exhibited the lowest calculation latency but suffered from clustered violations and risk underestimation. performance
Source sentence
EWMA exhibited superior performance in terms of calculation latency due to its computational simplicity, it fell short in adequately capturing the dynamics of volatility, resulting in clustered violations and underestimations of the risk level.
- The real‑time workflow delivers VaR estimates with latencies on the order of milliseconds. performance
Source sentence
This workflow yields reasonably robust VaR estimates with calculation latencies on the order of milliseconds.
- DCC‑GARCH estimator showed more profound underestimation compared to EWMA, as indicated by a larger number of violations across VaR levels. statistical
Source sentence
the issue of underestimation is more profound in DCC-GARCH estimator than EWMA estimator as indicated by the larger number of violations across the three VaR levels.
- HAR passed unconditional coverage and independence tests for all three VaR levels, while EWMA and GARCH were rejected at higher confidence levels. statistical
Source sentence
For HAR estimator, while it showed a robust performance for 95% 1‑day VaR metrics, it should be rejected at 99% confidence level for 97.5% and 99% 1‑day VaR.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 28 related papers). Coverage is partial, so this is neither replication nor refutation.
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.