Probabilistic Forecasting Cryptocurrencies Volatility: From Point to Quantile Forecasts

Grzegorz Dudek; Piotr Fiszeder; Witold Orzeszko

REVIEW 3 major objections 2 minor 1 cited by

Probabilistic Forecasting Cryptocurrencies Volatility: From Point to Quantile Forecasts

T0 review · 3 major / 2 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper argues that a simple residual-resampling method, QRS, applied to linear models of log-transformed realized volatility, produces the most accurate probabilistic forecasts of Bitcoin volatility among a wide field of statistical and

desk verdict The abstract describes a useful crypto-volatility forecasting benchmark, but the manuscript body is an unrelated plasticity paper, so there is nothing to review. read the letter →

arxiv 2508.15922 v1 pith:MHTGC7Y7 submitted 2025-08-21 q-fin.ST cs.AIcs.LG

Grzegorz Dudek , Witold Orzeszko , Piotr Fiszeder This is my paper · ORCID

classification q-fin.STcs.AIcs.LG

keywords probabilisticforecastingrealizedvolatilitycryptocurrencyBitcoinquantileregressionresidualsimulationmodelstacking

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that probabilistic forecasts of cryptocurrency realized variance can be built cheaply from ordinary point forecasts, and that the simplest route—simulating quantiles by resampling residuals from linear models of log-transformed realized volatility—is consistently more accurate than heavier statistical and machine-learning alternatives. The authors evaluate a wide range of base models, including HAR, GARCH, ARFIMA, LASSO, SVR, MLP, Random Forest, and LSTM, and wrap them in a probabilistic-stacking framework for Bitcoin. If the claim holds, traders and risk managers can obtain well-calibrated volatility quantiles without bespoke deep-learning architectures, and the bar for beating the linear-simulation baseline becomes clearer.

What carries the argument

QRS (Quantile Estimation through Residual Simulation): a residual-resampling procedure that turns a point forecast of log realized volatility into a set of simulated trajectories by adding draws from the model's in-sample residual distribution to the point forecast. Exponentiating back to realized-variance space and taking quantiles yields probabilistic forecasts. The method's success is tied to linear base models and the log transformation, which stabilizes the variance of the target series.

What would settle it

An out-of-sample test on a different cryptocurrency or a longer Bitcoin sample where QRS quantile coverage is systematically below the nominal level (e.g., the 5% quantile is exceeded more than 5% of the time) while a parametric GARCH quantile matches coverage would falsify the claim.

Watch

Extended reading notes

Core claim

The central claim is that the Quantile Estimation through Residual Simulation (QRS) method, applied to point forecasts from linear base models on log-transformed realized volatility, consistently outperforms more sophisticated alternatives for probabilistic forecasting of Bitcoin's realized variance. The QRS method converts a point forecast into a full conditional distribution by resampling the residuals of the point-forecast model and adding them to the forecast, thereby generating simulated paths of log realized volatility whose quantiles approximate those of the future realized variance. The paper reports this as the first systematic evaluation of such variance-quantile forecasts in crypt

Load-bearing premise

The method assumes that drawing residuals from the base model's historical errors reproduces the shape of the future conditional distribution of log realized volatility—that the error distribution is stable enough to resample.

Editorial extensions

If this is right

If QRS with linear bases is consistently best, then probabilistic volatility forecasting for Bitcoin can be done with transparent, reproducible models rather than opaque ML ensembles.
Residual simulation provides a distribution-free route to volatility quantiles, avoiding parametric GARCH distributional assumptions.
Probabilistic stacking of point-forecast-derived distributions inherits the strengths of the best base models, so gains come from input design (log realized volatility) rather than from the nonlinear learner.
The methodology can be applied directly to other cryptocurrencies or assets without retraining complex architectures.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural extension is to stress-test the QRS calibration on quantile coverage tests; the paper's reported superiority may be sensitive to the evaluation horizon and rolling-window length.
The same residual-simulation idea could be adapted to forecast tail-risk measures such as Value-at-Risk and Expected Shortfall for crypto portfolios, where extreme quantiles matter most.
Because the method is distribution-free, it may degrade less than parametric alternatives during volatility regime shifts, a hypothesis the paper's described experiments do not directly isolate.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague,

You need to know one thing up front: the submitted full text has nothing to do with the abstract. The abstract promises a systematic comparison of probabilistic volatility forecasts for Bitcoin using methods like QRS, HAR, GARCH, and LSTM. The body is a preprint on thermodynamically consistent neural yield functions for anisotropic plasticity. Same arXiv number? No. It looks like a file mix-up, but as it stands, the manuscript is internally incoherent.

What the abstract describes is actually a decent idea. A clean benchmark of quantile forecasting methods for crypto realized variance, with a simple residual-simulation approach against fancier ML alternatives, would be a useful contribution to crypto risk management. The "first systematic evaluation" claim is checkable and plausible, though not assured. The QRS method is essentially a residual bootstrap on point forecasts, which is standard but can be effective. If the real paper contains what the abstract says, it would be worth a careful read.

But none of that is in front of us. There is no derivation, no experimental protocol, no data, no results, no code. The soft spot is not a subtle assumption about residual resampling; it is the complete absence of any evidence for the claims. The stress-test note is correct: the document does not contain the study it purports to summarize. I cannot verify novelty, soundness, calibration, or even whether the authors actually ran the experiments. The reader's UNVERDICTED verdict is the only honest one.

I would not send this to peer review as is. A serious editor would desk reject or, more charitably, return it to the authors to correct the file. The abstract alone does not carry enough weight to justify referee time when the supporting text is missing. If a corrected manuscript arrives with the actual crypto content, then yes, I would send it out—the question is relevant and the proposed comparison is sensible.

For now: do not cite, do not discuss at reading group. Flag it as a submission error.

Best, [You]

Referee Report

3 major / 2 minor

Summary. The abstract of this submission claims a novel probabilistic forecasting framework for cryptocurrency realized variance, introducing the 'Quantile Estimation through Residual Simulation' (QRS) method, and reports that QRS applied to linear base models on log-transformed realized volatility outperforms more sophisticated alternatives for Bitcoin. The supplied full text, however, is a completely different preprint titled 'Thermodynamically Consistent Hybrid and Permutation-Invariant Neural Yield Functions for Anisotropic Plasticity' (arXiv:2508.15923v1). This body contains no mention of QRS, realized variance, Bitcoin, quantile forecasting, HAR/GARCH/ARFIMA, LASSO/SVR/MLP/LSTM, or any cryptocurrency data. The empirical results advertised in the abstract have no supporting derivation, experimental protocol, or evaluation in the manuscript text.

Significance. If the abstract's claims were supported, the paper would potentially offer a simple and practical baseline for probabilistic cryptocurrency volatility forecasting, with QRS as a lightweight alternative to fully conditional quantile models. The claimed robustness of probabilistic stacking would also be a useful practical contribution. However, because the manuscript body is an unrelated paper on plasticity, the claims cannot be assessed at all. There is no reproducible code, no machine-checked proof, and no falsifiable evidence in the submitted document. The significance of the result is therefore entirely unevaluated.

major comments (3)

[Full text (entire body after abstract)] The manuscript body is an unrelated preprint on neural yield functions for anisotropic plasticity (arXiv:2508.15923v1). It contains no discussion of QRS, realized variance, cryptocurrency volatility, Bitcoin, or any forecasting experiment. Consequently, the central claim of the abstract—that QRS applied to linear base models on log-transformed realized volatility outperforms alternatives—has no supporting evidence in the submitted document. This is not a subtle methodological issue; the evidential basis for the paper's stated contribution is entirely absent.
[Abstract vs Full text] The abstract describes a 'first study' proposing probabilistic forecasting of cryptocurrency variance, but the supplied full text does not present any such study. There are no equations, no data description, no evaluation metrics (e.g., quantile loss, coverage, Winkler score), and no comparison tables. The phrase 'consistently outperforms' is therefore an unsupported assertion. As a reviewer, I cannot verify any of the empirical claims, including the assumed validity of residual-simulation quantile estimation.
[Manuscript metadata] The document's running arXiv identifier (2508.15923) is different from the submission identifier (2508.15922), and the title, authors, and subject matter are entirely different. This mismatch further confirms that the abstract and body are not part of the same work. If this is an accidental file swap, the correct manuscript should be provided; as submitted, the paper cannot be evaluated for publication.

minor comments (2)

[General formatting] No page numbers, section numbers, or references are provided in the body, making it impossible to navigate or verify any specific claim. This is a significant presentation deficiency, though secondary to the substantive mismatch.
[Title and abstract] The title and abstract do not match the body. Even if the correct manuscript were uploaded, the authors should ensure that the abstract accurately reflects the content and that all arXiv identifiers are consistent.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular reasoning found; manuscript text is internally mismatched, making the abstract's QRS/volatility claims unsupported rather than circularly derived.

full rationale

The supplied full text is a different preprint on 'Thermodynamically Consistent Hybrid and Permutation-Invariant Neural Yield Functions for Anisotropic Plasticity' (arXiv:2508.15923), with no mention of QRS, realized volatility, Bitcoin, quantile forecasting, or any cryptocurrency data. The abstract's derivation chain for the QRS method therefore cannot be walked: there are no equations, no fitted parameters, no residual-simulation steps, and no empirical comparisons in the available body from which the abstract's claim could be reduced to its inputs. This is not a circularity — it is a complete evidential mismatch. Even if one considered the abstract alone, the described QRS approach (resampling residuals from point forecasts of log-realized volatility to estimate quantiles) is a standard residual-bootstrap technique and is not definitionally circular: the quantiles are conditional on the point-forecast model and are evaluated against held-out realized variance, so no direct equation-to-equation reduction is apparent. The central problem is that the submitted manuscript does not contain the study the abstract purports to report. That is a serious integrity and reproducibility concern, but it falls outside the seven enumerated circularity patterns. Accordingly, the circularity score is 0, with no circular steps identified.

Assumptions & free parameters 0 free parameters · 1 assumptions · 0 invented entities

No free parameters are identifiable from the abstract. The QRS method relies on a distributional assumption about residuals, listed above. No new entities are introduced.

assumptions (1)

domain assumption Residual simulation from point forecasts approximates the conditional distribution of realized variance.
The abstract's QRS method rests on the idea that resampling residuals from base model point forecasts yields valid quantile estimates. The full text does not test this, and the abstract does not state conditions.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Probabilistic Forecasting Cryptocurrencies Volatility: From Point to Quantile Forecasts." pith.science (2026). https://pith.science/paper/MHTGC7Y7

@misc{pith2026250815922,
  author       = {Pith},
  title        = {Pith review of: Probabilistic Forecasting Cryptocurrencies Volatility: From Point to Quantile Forecasts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MHTGC7Y7}},
  note         = {Machine review of arXiv:2508.15922}
}

read the original abstract

Cryptocurrency markets are characterized by extreme volatility, making accurate forecasts essential for effective risk management and informed trading strategies. Traditional deterministic (point) forecasting methods are inadequate for capturing the full spectrum of potential volatility outcomes, underscoring the importance of probabilistic approaches. To address this limitation, this paper introduces probabilistic forecasting methods that leverage point forecasts from a wide range of base models, including statistical (HAR, GARCH, ARFIMA) and machine learning (e.g. LASSO, SVR, MLP, Random Forest, LSTM) algorithms, to estimate conditional quantiles of cryptocurrency realized variance. To the best of our knowledge, this is the first study in the literature to propose and systematically evaluate probabilistic forecasts of variance in cryptocurrency markets based on predictions derived from multiple base models. Our empirical results for Bitcoin demonstrate that the Quantile Estimation through Residual Simulation (QRS) method, particularly when applied to linear base models operating on log-transformed realized volatility data, consistently outperforms more sophisticated alternatives. Additionally, we highlight the robustness of the probabilistic stacking framework, providing comprehensive insights into uncertainty and risk inherent in cryptocurrency volatility forecasting. This research fills a significant gap in the literature, contributing practical probabilistic forecasting methodologies tailored specifically to cryptocurrency markets.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Fermi-Large Area Telescope Detection of Very High Energy (>100 GeV) Emission from Misaligned Jetted Active Galactic Nuclei
astro-ph.HE 2025-08 conditional novelty 4.0 of 10

A systematic Fermi-LAT search of 160 radio galaxies yields 12 firm detections above 100 GeV, including two new very high energy radio galaxies.

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages · cited by 1 Pith paper

[1]

Jadoon, Department of Aerospace Engineering & Engineering Mechanics, The University of Texas at Austin, Austin, TX 78712 Ravi G

THERMODYNAMICALLY CONSISTENT HYBRID AND PERMUTATION -I NVARIANT NEURAL YIELD FUNCTIONS FOR ANISOTROPIC PLASTICITY A PREPRINT Asghar A. Jadoon, Department of Aerospace Engineering & Engineering Mechanics, The University of Texas at Austin, Austin, TX 78712 Ravi G. Patel, Sandia National Laboratories, Albuquerque, NM 87185 Brian N. Granzow, Sandia National ...

work page 2025
[2]

Thermodynamically Consistent Hybrid and Permutation-Invariant Neural Yield Functions for Anisotropic Plasticity

linear stress transformations. We calibrate the proposed frameworks on a sparse Al-7079 extrusion experimental dataset compris- ing 12 uniaxial samples with measured yield stresses and Lankford ratios. To test the robustness of each framework, nine datasets were generated using k-fold cross-validation. These datasets were then used to quantitatively compa...

work page Pith review arXiv 2025

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

Jadoon, Department of Aerospace Engineering & Engineering Mechanics, The University of Texas at Austin, Austin, TX 78712 Ravi G

THERMODYNAMICALLY CONSISTENT HYBRID AND PERMUTATION -I NVARIANT NEURAL YIELD FUNCTIONS FOR ANISOTROPIC PLASTICITY A PREPRINT Asghar A. Jadoon, Department of Aerospace Engineering & Engineering Mechanics, The University of Texas at Austin, Austin, TX 78712 Ravi G. Patel, Sandia National Laboratories, Albuquerque, NM 87185 Brian N. Granzow, Sandia National ...

work page 2025

[2] [2]

Thermodynamically Consistent Hybrid and Permutation-Invariant Neural Yield Functions for Anisotropic Plasticity

linear stress transformations. We calibrate the proposed frameworks on a sparse Al-7079 extrusion experimental dataset compris- ing 12 uniaxial samples with measured yield stresses and Lankford ratios. To test the robustness of each framework, nine datasets were generated using k-fold cross-validation. These datasets were then used to quantitatively compa...

work page Pith review arXiv 2025