Skip to content

2020 Q4 Quarterly Research Archive

Records scoring at least 40 within the primary scope pass rule review and are published without additional manual review. This page does not validate author claims or provide investment advice.

  • Coverage: 2020-10-01 to 2020-12-31
  • Passed rule review: 11
  • Sources: 6

Topic distribution

Domains

  • volatility: 9
  • hedging exposure risk: 2
  • execution costs: 2
  • lifecycle infrastructure: 1

Methods

  • research methods: 6
  • financial ml: 5

Facets

  • instrument vix options: 1

Passed rule review

The term structure of the VXX option smirk: Pricing VXX option with a two‐factor model and asymmetry jumps

  • Published: 2020-12-28
  • Source: Journal of Futures Markets
  • Publication status: peer_reviewed
  • Original source: Open original source

Why it matters

The model uses jump-to-default and asymmetric jumps to describe VXX option smirk and term structure and reports better fit within quoted spreads; model fit is not the same as an executable pricing advantage. (abstract:S1, abstract:S2, abstract:S4)

Main author claims

  • The authors claim that their model outperforms Bao et al.'s model by 28.19% in-sample and 23.38% out-of-sample. (abstract:S3)
  • The authors claim that the model improves the probability that estimated prices fall inside the quoted bid-ask spread and better fits the term structure of VXX implied volatility, especially for out-of-the-money options. (abstract:S4)
  • The authors claim that the model provides a more flexible framework for capturing the time variation in VXX options smirk and volatility term structure. (abstract:S2)

Data, method, or discussion scope

The study uses market data for VXX options empirically, compares with existing models, and evaluates the proportion of prices within bid-ask spreads. (abstract:S3, abstract:S4)

Main limitations

The model is tailored to VXX options; generalization to other volatility products requires additional verification; model calibration complexity and computation time are not discussed; the sample period is not specified in the abstract. (abstract:S1, abstract:S3)

A short cut: Directly pricing VIX futures with discrete‐time long memory model and asymmetric jumps

  • Published: 2020-12-28
  • Source: Journal of Futures Markets
  • Publication status: peer_reviewed
  • Original source: Open original source

Why it matters

The model prices VIX futures directly from VIX dynamics and uses an analytic form to reduce numerical-integration burden; stable error reductions across regimes still require full-result scrutiny. (abstract:S1, abstract:S2, abstract:S3, abstract:S4)

Main author claims

  • The authors claim that their model significantly reduces pricing errors compared to existing models using realized variance, both in- and out-of-sample. (abstract:S4)
  • The authors claim that the framework has fewer parameter constraints and an analytical solution, avoiding time-consuming and sometimes unstable numerical integration. (abstract:S2, abstract:S3)
  • The authors claim that besides seeking better current volatility measures, it is also important to utilize information embedded in the VIX itself. (abstract:S5)

Data, method, or discussion scope

The study uses market data for VIX futures and compares pricing errors in- and out-of-sample with models based on realized variance. (abstract:S4)

Main limitations

The model applies only to VIX futures; extension to other volatility products is not discussed; it relies on the representativeness of VIX historical data; market liquidity or microstructure effects are not mentioned. (abstract:S1, abstract:S4)

The Deep Parametric PDE Method: Application to Option Pricing

  • Published: 2020-12-11
  • Source: arXiv Quantitative Finance History
  • Publication status: preprint
  • Original source: Open original source

Why it matters

Approximating a family of high-dimensional parametric PDEs with one network could reduce repeated-pricing latency; the reported millisecond and 25-dimensional results do not establish full production-stack accuracy or cost bounds. (abstract:S1, abstract:S2, abstract:S4)

Main author claims

  • The authors claim that the deep parametric PDE method can solve high-dimensional parametric PDEs without the need of sample solutions during training. (abstract:S1, abstract:S2)
  • The authors claim that after a single training phase, option prices for various times, states, and model parameters are available in milliseconds. (abstract:S4)
  • The authors claim that accuracy in price and implied volatility generalization is validated up to 25 dimensions, and it outperforms alternative machine learning approaches. (abstract:S5, abstract:S6)

Data, method, or discussion scope

The study uses numerical experiments in the multivariate Black-Scholes model up to 25 dimensions, comparing with other machine learning methods. (abstract:S3, abstract:S5, abstract:S6)

Main limitations

Experiments are limited to the Black-Scholes framework, possibly not covering more complex stochastic processes; deep learning methods may still face residual curse of dimensionality; calibration to real market data is not mentioned. (abstract:S3, abstract:S5)

Forecasting VIX Using Filtered Historical Simulation

  • Published: 2020-12-06
  • Source: Journal of Financial Econometrics
  • Publication status: peer_reviewed
  • Original source: Open original source

Why it matters

The method combines filtered historical simulation with GARCH to address skew and heavy tails in VIX and reports gains over a benchmark; the abstract cannot establish regime robustness or trading value. (abstract:S1, abstract:S2, abstract:S4, abstract:S5)

Main author claims

  • The paper reports lower forecast errors than the Hao–Zhang Normal-VIX benchmark in both fitted and held-out evaluations. (abstract:S4)
  • The authors claim that using volatility indices instead of the options-based pricing method significantly reduces computational burden. (abstract:S5)
  • The authors claim that the method accommodates non-normalities such as negative skewness and positive excess kurtosis. (abstract:S2)

Data, method, or discussion scope

The study uses four established volatility indices for in-sample and out-of-sample evaluation and comparison with a benchmark model. (abstract:S3, abstract:S4)

Main limitations

Comparison is limited to a single benchmark model, possibly omitting other competitive methods; robustness to outdated data or structural breaks is not mentioned; the journal is peer-reviewed but the abstract does not mention the data period. (abstract:S4)

Moment Risk Premia and Stock Return Predictability

  • Published: 2020-11-26
  • Source: Journal of Financial and Quantitative Analysis
  • Publication status: peer_reviewed
  • Original source: Open original source

Why it matters

The authors report incremental predictive information in higher-moment risk premia across horizons, potentially relevant to horizon-specific risk budgets; allocation value still depends on construction, costs, and constraints. (abstract:S1, abstract:S2, abstract:S3, abstract:S4)

Main author claims

  • The authors claim that the second-moment risk premium predicts market returns at short horizons with positive coefficients, while the third-moment premium predicts at medium horizons with negative coefficients, and the fourth-moment with positive coefficients. (abstract:S2)
  • The authors claim that combining higher-moment premia improves stock return predictability both in-sample and out-of-sample. (abstract:S3)
  • The paper reports economic relevance in its allocation exercise and says the result remains under its robustness checks. (abstract:S4)

Data, method, or discussion scope

The study is based on equity market data and option-implied moment risk premia, using in-sample and out-of-sample prediction evaluation, an asset-allocation exercise, and robustness checks. (abstract:S2, abstract:S3, abstract:S4)

Main limitations

The horizon-dependence of predictive coefficient signs may add complexity in practical use; the scope of robustness checks is not detailed; performance in other asset classes is not mentioned. (abstract:S2, abstract:S4)

The Mysteries and Makings of Machine Learning with Dr. Ernie Chan of QTS Cap

  • Published: 2020-11-19
  • Source: The Derivative by RCM Alternatives
  • Publication status: unknown
  • Original source: Open original source

Why it matters

The podcast is useful for seeing how a practitioner frames ML models, strategy choice, and durable edge, but its description supplies topic context rather than model or performance evidence. (description:S1, description:S4, description:S5, description:S6)

Main author claims

  • The authors report: The episode description claims that the discussion will cover the practical effectiveness of machine learning in investing, including whether it works and why not all AI funds succeed. (description:S1, description:S2)
  • The authors report: The episode description claims that through guest Ernie Chan's background, topics such as decision trees, ensemble methods, and random forest techniques will be explored. (description:S5, description:S6)
  • The authors report: The episode description implies that it is still possible to gain an edge with machine learning, as indicated by the chapter title 'You can Still Gain an Edge'. (description:S7)

Data, method, or discussion scope

This entry is a podcast description; it contains no research evidence. The content is a discussion of guest opinions, without systematic empirical analysis. (description:S6, description:S12)

Main limitations

The podcast content is opinion-based and not peer-reviewed; the guest's affiliated firm may have conflicts of interest; the performance track records provided do not guarantee future results. (description:S12, description:S14, description:S15)

Deep Smoothing of the Implied Volatility Surface

  • Published: 2020-10-26
  • Source: arXiv Quantitative Finance History
  • Publication status: preprint
  • Original source: Open original source

Why it matters

The method jointly addresses IV-surface fit, structural constraints, and uncertainty in sparse regions, all relevant to production models; soft penalties, however, do not by themselves guarantee strict no-arbitrage. (abstract:S2, abstract:S3, abstract:S5, abstract:S9)

Main author claims

  • The authors claim that their neural network approach guarantees no arbitrage by penalizing the loss with soft constraints. (abstract:S5)
  • The authors claim that the method can serve as a plug-in correction for standard IVS models when they fail to replicate market prices. (abstract:S6, abstract:S7)
  • The authors claim that deeper neural networks improve performance and that the approach is particularly useful when data are sparse or erroneous. (abstract:S8, abstract:S10)

Data, method, or discussion scope

The evidence is based on empirical evaluation on training and testing sets, benchmarking against standard IVS models, and testing under varying data quality. (abstract:S11, abstract:S12, abstract:S8)

Main limitations

Soft constraints may not strictly guarantee no-arbitrage in all scenarios; the method relies on standard IVS models as a base; performance under extreme market conditions is not discussed. (abstract:S5, abstract:S6, abstract:S7)

Option Hedging with Risk Averse Reinforcement Learning

  • Published: 2020-10-23
  • Source: arXiv Quantitative Finance History
  • Publication status: preprint
  • Original source: Open original source

Why it matters

This preprint places risk aversion, hedge volatility, and transaction costs in one reinforcement-learning framework, making it a candidate approach for cost-aware hedging; its comparative performance remains an abstract-level claim. (abstract:S1, abstract:S2, abstract:S5)

Main author claims

  • The authors claim that risk-averse reinforcement learning using the TRVO algorithm can be applied to option hedging. (abstract:S1, abstract:S2)
  • The authors claim that by training a sheaf of agents with different risk aversions, an efficient frontier in the volatility-P&L space can be spanned. (abstract:S4)
  • The authors claim that the derived hedging strategy outperforms the Black-Scholes delta hedge and is robust and flexible across different market behaviors. (abstract:S5)

Data, method, or discussion scope

The study is conducted in a simulated vanilla option hedging environment with discrete time and transaction costs, using a sheaf of agents with varying risk appetites. The evidence is based on simulation experiments; no real market data is mentioned. (abstract:S2, abstract:S3, abstract:S4, abstract:S5)

Main limitations

The authors do not provide validation on real market data; the simulated environment may oversimplify market frictions and model risk; the paper is a preprint and has not been peer-reviewed. (abstract:S2, abstract:S5)

Deep Learning-Based Least Square Forward-Backward Stochastic Differential Equation Solver for High-Dimensional Derivative Pricing

  • Published: 2020-10-12
  • Source: arXiv Quantitative Finance History
  • Publication status: preprint
  • Original source: Open original source

Why it matters

The method combines deep learning with least-squares regression for high-dimensional derivative pricing, offering a candidate numerical route; the abstract does not define dimensional, error, convergence, or compute-cost bounds. (abstract:S1, abstract:S2, abstract:S3)

Main author claims

  • The authors report: A new forward-backward stochastic differential equation solver is proposed, combining a deep learning solver with the least squares Monte Carlo technique for American option valuation. (abstract:S1)
  • The authors report: Numerical experiments demonstrate the efficiency and accuracy of the solver for pricing complex early exercise derivatives such as callable yield notes. (abstract:S2)
  • The authors report: The method can serve as a generic numerical solver for pricing derivatives across various asset groups, particularly for high-dimensional derivatives with early exercise features. (abstract:S3)

Data, method, or discussion scope

The study tests the proposed solver via numerical experiments on high-dimensional derivatives like callable yield notes, though specific dimensions and benchmark models are not disclosed. (abstract:S2)

Main limitations

The abstract lacks computational speed comparisons, a clear definition of the high-dimensional range, model assumptions (e.g., the form of stochastic processes), and convergence analysis. (abstract:S1)

The implied volatility smirk of commodity options

  • Published: 2020-10-09
  • Source: Journal of Futures Markets
  • Publication status: peer_reviewed
  • Original source: Open original source

Why it matters

The authors report relations between commodity-option IV smirks and subsequent commodity and S&P 500 returns, making them candidate cross-asset predictors; the abstract does not establish tradability, causality, or data-mining robustness. (abstract:S4, abstract:S5)

Main author claims

  • The authors report: Commodity implied volatility curves are generally negatively skewed with a positive curvature. (abstract:S3)
  • The authors report: The information embedded in IV smirks can significantly predict monthly commodity and S&P500 returns. (abstract:S4)
  • The authors report: The risk-neutral fourth cumulant (FC) from the crude oil market outperforms all standard predictors for S&P500 returns. (abstract:S5)

Data, method, or discussion scope

The study adopts Zhang and Xiang's methodology, analyzes the term structure and dynamics of IV smirks across four commodity markets, and conducts in-sample and out-of-sample predictability tests. (abstract:S1, abstract:S2, abstract:S4)

Main limitations

Only four commodity markets are examined, and the prediction horizon is monthly; no discussion of data mining or robustness across economic cycles is provided. (abstract:S1, abstract:S4)

A Horserace of Volatility Models for Cryptocurrency: Evidence from Bitcoin Spot and Option Markets

  • Published: 2020-10-04
  • Source: arXiv Quantitative Finance History
  • Publication status: preprint
  • Original source: Open original source

Why it matters

The study compares Bitcoin volatility models and reports an option strategy based on forecast-versus-implied volatility spreads; without costs and execution details, this remains a research claim to be checked. (abstract:S3, abstract:S5, abstract:S6)

Main author claims

  • The authors report: GARCH and EGARCH models perform much better than other volatility models for Bitcoin spot price series. (abstract:S3)
  • The authors report: Bitcoin prices lack asymmetric volatility response to past returns, as the EGARCH asymmetric term is positive and insignificant. (abstract:S4)
  • The authors report: A simple option trading strategy exploiting the volatility spread between GARCH forecast and implied volatility, with delta-hedging, can yield robust profits. (abstract:S5, abstract:S6)

Data, method, or discussion scope

The study uses Bitcoin spot price series to test various volatility models, compares in-sample fit and out-of-sample forecasts, and formulates a volatility-spread trading strategy. (abstract:S1, abstract:S2, abstract:S5)

Main limitations

The abstract does not specify the time period, data frequency, option data source, or account for transaction costs; the profit robustness may be sensitive to strategy parameters. (abstract:S5, abstract:S6)