2021 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: 2021-10-01 to 2021-12-31
- Passed rule review: 12
- Sources: 5
Topic distribution¶
Domains¶
- volatility: 11
- hedging exposure risk: 2
- microstructure: 1
Methods¶
- research methods: 8
- financial ml: 5
Facets¶
- instrument vix options: 4
- instrument index options: 3
Passed rule review¶
Multistep forecast of the implied volatility surface using deep learning¶
- Published: 2021-12-27
- Source: Journal of Futures Markets
- Publication status:
peer_reviewed - Original source: Open original source
Why it matters¶
Multivariate and multi-step implied-volatility-surface forecasting is operationally relevant for option-pricing workflows, with LSTM and ConvLSTM compared in-sample and out-of-sample. (abstract:S1, abstract:S2, abstract:S3, abstract:S4)
Main author claims¶
- They model the full IV surface with LSTM and ConvLSTM to produce multivariate multi-step forecasts. (
abstract:S2) - The authors report: Using daily SPX options from 2002–2019, training fit MAPE is 3.56% for LSTM and 3.88% for ConvLSTM. (
abstract:S3) - The authors report: Out-of-sample, ConvLSTM (8.26% MAPE) significantly outperforms LSTM and traditional time-series models. (
abstract:S4)
Data, method, or discussion scope¶
Scope is SPX IV surface forecasting on daily options 2002–2019, comparing LSTM and ConvLSTM for in-sample fit and out-of-sample performance. (abstract:S2, abstract:S3, abstract:S4)
Main limitations¶
Network architecture details, validation split, update cadence, and deployment latency are not provided. (abstract:S3, abstract:S4)
Neural Networks for Delta Hedging¶
- Published: 2021-12-19
- Source: arXiv Quantitative Finance History
- Publication status:
preprint - Original source: Open original source
Why it matters¶
The study contrasts Black-Scholes ideal-market assumptions with real trading frictions and explores multiple neural hedging architectures for practical assessment. (abstract:S1, abstract:S2, abstract:S4, abstract:S5, abstract:S6)
Main author claims¶
- They state BS assumptions of zero transaction costs and continuous trading are difficult to satisfy in practice. (
abstract:S2) - They test hedging capacity across recurrent, temporal convolutional, attention, and Span MLP networks. (
abstract:S4) - They attempt combinations of traditional derivative-hedging models with DNN approaches and construct an NNHedge pipeline for model development and assessment. (
abstract:S5,abstract:S6)
Data, method, or discussion scope¶
Scope is limited to the abstract-described hedging architecture comparison and the proposed NNHedge framework. (abstract:S4, abstract:S5, abstract:S6)
Main limitations¶
No assets, backtest protocol, error metrics, or transaction-cost model are provided to operationalize the claimed improvements. (abstract:S1, abstract:S4, abstract:S5)
Multivariate Realized Volatility Forecasting with Graph Neural Network¶
- Published: 2021-12-17
- Source: arXiv Quantitative Finance History
- Publication status:
preprint - Original source: Open original source
Why it matters¶
Integrating relational structure with limit-order-book data may improve multivariate short-term volatility forecasts; the abstract does not show that these forecasts translate into tradable signals. (abstract:S1, abstract:S2, abstract:S3, abstract:S4, abstract:S5, abstract:S6)
Main author claims¶
- They introduce a Graph Transformer Network for short-term realized-volatility forecasting in a multivariate, order-book-based framework. (
abstract:S3,abstract:S4) - The authors report: The model combines limit order book features with temporal and cross-sectional relations from multiple sources. (
abstract:S5) - The authors report: Experiments on about 500 stocks from the S&P 500 report better performance than benchmarks. (
abstract:S6)
Data, method, or discussion scope¶
Scope is multivariate short-horizon realized-volatility forecasting over roughly 500 S&P 500 stocks using graph-transformer models and LO B inputs. (abstract:S3, abstract:S5, abstract:S6)
Main limitations¶
Graph construction, benchmark definitions, and computational constraints are not specified, and external generalization checks are absent. (abstract:S4, abstract:S5, abstract:S6)
Increasing the information content of realized volatility forecasts¶
- Published: 2021-12-11
- Source: Journal of Financial Econometrics
- Publication status:
peer_reviewed - Original source: Open original source
Why it matters¶
The paper proposes replacing the Calendar RV estimator with Rolling RV to increase usable observations in out-of-sample realized-volatility forecasting. (abstract:S1, abstract:S2, abstract:S3, abstract:S4)
Main author claims¶
- They contrast Calendar RV with a Rolling RV that sums trailing M returns at each timestamp. (
abstract:S1,abstract:S2) - The authors report: In out-of-sample one-day forecasting, rolling RV uses more datapoints and is claimed to potentially reduce standard errors. (
abstract:S3) - The authors report: For S&P 500 and 26 DJIA stocks, the Rolling approach is reported to generally yield statistically and economically significant superior out-of-sample performance. (
abstract:S4)
Data, method, or discussion scope¶
The scope is out-of-sample realized-volatility model evaluation on S&P 500 and 26 DJIA stocks comparing Calendar and Rolling RV. (abstract:S3, abstract:S4)
Main limitations¶
No full regression setup, sampling frequency, missing-data handling, significance criteria, or mapping to trading-cost-aware performance is provided. (abstract:S3, abstract:S4)
Risk of Bitcoin Market: Volatility, Jumps, and Forecasts¶
- Published: 2021-12-09
- Source: arXiv Quantitative Finance History
- Publication status:
preprint - Original source: Open original source
Why it matters¶
The abstract centers risk dynamics of Bitcoin on realized volatility and jumps, which affect risk management and derivatives modeling under volatile crypto conditions. (abstract:S4, abstract:S5, abstract:S6, abstract:S7)
Main author claims¶
- The authors claim the Bitcoin market is extremely risky in terms of volatility and entangled, extensive consecutive jumps. (
abstract:S5) - The authors report: Empirical results are reported that lagged realized variance increases future realized variance, while jumps, especially positive ones, reduce it. (
abstract:S6) - The authors report: For long-horizon realized-variance forecasting, explicitly modeling jumps and signed estimators is claimed to improve forecast accuracy and utility, while this is unnecessary for short-term forecasts. (
abstract:S7)
Data, method, or discussion scope¶
Scope is realized-volatility risk dynamics in Bitcoin, focusing on lagged RV and jump effects on future realized variance. (abstract:S4, abstract:S6, abstract:S7)
Main limitations¶
The abstract gives directional findings only and omits sample windows, event taxonomy, jump-identification and estimation details, and the precise definition of major incidents. (abstract:S5, abstract:S6, abstract:S7)
Forecast Evaluation in Large Cross-Sections of Realized Volatility¶
- Published: 2021-12-09
- Source: arXiv Quantitative Finance History
- Publication status:
preprint - Original source: Open original source
Why it matters¶
It introduces cross-section-aware equal-predictive-accuracy testing for realized volatility, relevant to model benchmarking when assets are interdependent. (abstract:S1, abstract:S2, abstract:S3, abstract:S4, abstract:S5)
Main author claims¶
- The paper addresses forecast evaluation under cross-sectional dependence for realized volatility measures. (
abstract:S1,abstract:S2) - The authors report: Under equal predictive accuracy null, standard HAR is benchmark; under unequal predictive accuracy alternative, an augmented HAR estimated via LASSO is compared. (
abstract:S3) - The study adds measurement-error correction and cross-sectional jump components for sensitivity, with out-of-sample assessment via numerical implementation. (
abstract:S4,abstract:S5)
Data, method, or discussion scope¶
Scope is forecast evaluation procedures for realized volatility under cross-sectional dependence, including HAR vs augmented HAR-LASSO and numerical out-of-sample implementations. (abstract:S1, abstract:S3, abstract:S4, abstract:S5)
Main limitations¶
No asset universe size, significance thresholds, and concrete error metrics for model comparison are specified. (abstract:S1, abstract:S2, abstract:S5)
Volatility model applications in China's SSE50 options market¶
- Published: 2021-12-07
- Source: Journal of Futures Markets
- Publication status:
peer_reviewed - Original source: Open original source
Why it matters¶
The study compares volatility models for SSE50 options, including GARCH family performance and implied-volatility spread strategy implications. (abstract:S1, abstract:S2, abstract:S3, abstract:S4, abstract:S5)
Main author claims¶
- The authors report: GARCH and its variants are reported to outperform ARCH in in-sample fitting. (
abstract:S2) - The authors report: In five of seven options, GARCH volatility forecasts outperform implied volatility for realized-volatility forecasting out of sample. (
abstract:S4) - The authors report: A trading strategy is formulated from the GARCH-implied-volatility spread and described as robustly profitable. (
abstract:S5)
Data, method, or discussion scope¶
In-sample and out-of-sample model comparison on SSE50 options, with a spread-based strategy proposal. (abstract:S1, abstract:S2, abstract:S4, abstract:S5)
Main limitations¶
Sample definitions, trading-cost assumptions, and risk-adjusted performance definitions are absent despite claims about robust profitability. (abstract:S4, abstract:S5)
Reinforcement learning for options on target volatility funds¶
- Published: 2021-12-03
- Source: arXiv Quantitative Finance History
- Publication status:
preprint - Original source: Open original source
Why it matters¶
By framing TVS option pricing with funding-cost-aware control, the paper links closed-form BS solutions and reinforcement-learning approaches for cases without analytic solutions. (abstract:S1, abstract:S2, abstract:S3, abstract:S4, abstract:S5)
Main author claims¶
- The authors report: The abstract frames option pricing for target volatility strategies as a control problem with funding costs and component-hedging-cost heterogeneity. (
abstract:S1,abstract:S2) - The authors report: An analytical solution is derived in the BS setting; RL is then used under local-volatility to obtain conservative pricing under uncertainty. (
abstract:S3,abstract:S4) - The authors report: RL agent performance is reported to be compatible with path-wise BS analytical strategy results, indicating competitiveness in the LV setting. (
abstract:S5)
Data, method, or discussion scope¶
Pricing-control formulation for TVS option pricing under BS and local-volatility dynamics, with RL-based comparison. (abstract:S1, abstract:S3, abstract:S4, abstract:S5)
Main limitations¶
The abstract omits RL training details, computational scale, convergence criteria, and implementation constraints under real trading frictions. (abstract:S4, abstract:S5)
Data-driven Hedging of Stock Index Options via Deep Learning¶
- Published: 2021-11-05
- Source: arXiv Quantitative Finance History
- Publication status:
preprint - Original source: Open original source
Why it matters¶
Learning hedge ratios directly from option data, with sentiment inputs, is relevant to data-driven hedging design and the limits of model-only hedging. (abstract:S1, abstract:S2, abstract:S3, abstract:S4)
Main author claims¶
- The paper develops deep-learning models to learn hedge ratios directly from S&P 500 options data. (
abstract:S1) - The authors report: A feedforward network using maturity, Black-Scholes delta, and sentiment variables is reported to perform best in out-of-sample tests. (
abstract:S2) - The authors report: The model is claimed to outperform standard BS delta hedging and a recent data-driven benchmark, highlighting sentiment importance. (
abstract:S3,abstract:S4)
Data, method, or discussion scope¶
Out-of-sample comparison of deep-learning feature combinations for hedge-ratio learning on S&P 500 options data. (abstract:S1, abstract:S2, abstract:S3, abstract:S4)
Main limitations¶
No transaction-cost assumptions, hedging interval, risk metric definitions, or benchmark details are provided. (abstract:S2, abstract:S3, abstract:S4)
Convex (CONVX) to the Core with Certeza’s Brett Nelson¶
- Published: 2021-10-28
- Source: The Derivative by RCM Alternatives
- Publication status:
unknown - Original source: Open original source
Why it matters¶
The episode offers practitioner views on Vega/Gamma and regime framing, useful context for strategy framing but not a validated empirical result source. (description:S1, description:S3, description:S4, description:S6, description:S11, description:S12)
Main author claims¶
- The authors report: The episode discusses Certeza’s Convex Core fund and how it relates to their macro Volatility strategy. (
description:S1,description:S2) - The authors report: Topics include Vega and Gamma usage, flow importance, and hedging behavior across market regimes. (
description:S3,description:S4,description:S6) - The authors report: The description includes legal/usage disclaimers that content is informational only and not legal, business, or tax advice. (
description:S11,description:S12,description:S13)
Data, method, or discussion scope¶
Scope is limited to episode-level discussion topics and disclosures, not empirical or reproducible trading tests. (description:S1, description:S6, description:S9)
Main limitations¶
Missing verifiable metrics, position parameters, execution-cost assumptions, and risk statistics means strategy efficacy cannot be validated from this source. (description:S1, description:S9, description:S10)
Do VIX futures contribute to the valuation of VIX options?¶
- Published: 2021-10-18
- Source: Journal of Futures Markets
- Publication status:
peer_reviewed - Original source: Open original source
Why it matters¶
The paper evaluates whether including VIX futures as state variables materially improves VIX options pricing relative to common alternatives. (abstract:S1, abstract:S2, abstract:S3, abstract:S4, abstract:S5)
Main author claims¶
- The authors report: The abstract states VIX futures are used for hedging nontradable VIX exposure, but their information role in option pricing is not fully explored. (
abstract:S1,abstract:S2) - The authors report: Two types of VIX option pricing formulas are derived, using VIX index or VIX futures state variables in a discrete-time setting. (
abstract:S3) - The authors report: Models using VIX futures are claimed to significantly outperform models based on SPX returns, realized volatility, or the VIX index. (
abstract:S4)
Data, method, or discussion scope¶
The evidence concerns alternative VIX-option pricing specifications with simple discrete-time dynamics and out-of-sample robustness checks. (abstract:S3, abstract:S4, abstract:S5)
Main limitations¶
No details on parameter calibration, option sample composition, or exact robustness criteria are included. (abstract:S3, abstract:S4, abstract:S5)
An Application of Damped Diffusion for Modeling Volatility Dynamics¶
- Published: 2021-10-18
- Source: Journal of Financial Econometrics
- Publication status:
peer_reviewed - Original source: Open original source
Why it matters¶
A damped CEV volatility model is proposed with tractable inference and joint physical/risk-neutral estimation, which affects both variance forecasting and option-pricing pipelines. (abstract:S1, abstract:S2, abstract:S3, abstract:S4, abstract:S5)
Main author claims¶
- The authors report: The DCEV model is introduced to remedy explosive behavior in CEV and to better accommodate mean reversion. (
abstract:S1) - The authors report: With linear drift, the model allows analytic inference of latent variances from VIX, followed by maximum-likelihood estimation under both physical and risk-neutral measures. (
abstract:S2) - The authors report: The abstract claims superior in-sample fit, out-of-sample physical variance forecasting, and out-of-sample option-pricing performance versus CEV/NLD/Heston benchmarks. (
abstract:S3,abstract:S4)
Data, method, or discussion scope¶
DCEV estimation and performance comparisons using S&P 500 returns and inferred variances across physical and risk-neutral settings. (abstract:S2, abstract:S3, abstract:S4)
Main limitations¶
Performance claims are high-level; no forecast window, sample frequency, parameter constraints, or misspecification stress tests are provided. (abstract:S3, abstract:S4, abstract:S5)