Pith. sign in

REVIEW 4 major objections 6 minor 66 references

Simple and Effective Portfolio Construction with Crypto Assets

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A fixed 90/10 traditional-and-crypto allocation, diluted with cash to a target risk, matches a full risk-allocation optimizer in backtests, with Sharpe rising from 0.73 to 1.06.

desk verdict Clean, honest practitioner note with standard math and a useful empirical simplification; the headline Sharpe gap is in-sample-selected and cost-blind, so it needs a walk-forward and a transaction-cost model before I'd trust it. read the letter →

arxiv 2412.02654 v1 pith:CUPM6NX2 submitted 2024-12-03 econ.EM

Kasper Johansson , Stephen Boyd This is my paper · ORCID
classification econ.EM MSC 91G1090C25
keywords portfolioconstructioncryptoassetsriskparityconstrainedallocationdynamiccashdilutionconvexoptimizationSharperatio
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

The paper argues that crypto assets can be added to a conventional portfolio with ordinary risk-allocation tools, despite their extreme volatility and heavy tails. Its central demonstration is a constrained risk-allocation (CRA) portfolio that sets risk contributions and scales exposure to a 10% annualized risk cap; adding 10% crypto raises the backtested Sharpe ratio from 0.73 for an industry-only portfolio to 1.00, with similar volatility and drawdown. Inspecting those weights suggests an even simpler rule, DD90/10: hold 90% traditional industry portfolios and 10% crypto in fixed relative weights, and dilute with cash each day to hit the same risk target. In the same backtest this rule reports Sharpe 1.06 with an EWMA volatility estimate and 1.04 with GARCH, slightly better than the optimizer. The practical point is that investors may not need return forecasts or machine-learning models to benefit from crypto exposure.

What carries the argument

The central object is the constrained risk-allocation (CRA) problem: choose weights $w \ge 0$ and cash $c$ to minimize cash holdings subject to risk contributions $w_i(\Sigma w)_i = \rho_i w^T\Sigma w$, a risk cap $w^T\Sigma w \le \sigma^2$, and weight constraints. Its solution rests on the fact that the risk-allocation equations force $w = \alpha x^\star$, where $x^\star$ is the unique minimizer of $(1/2)x^T\Sigma x - \sum_i \rho_i \log x_i$, so the full problem collapses to a scalar scaling rule, $\alpha^\star = \min\{1/(\mathbf{1}^T x^\star),\, \sigma/\sqrt{(x^\star)^T\Sigma x^\star},\, g_i/(F x^\star)_i\}$. The takeaway rule DD90/10 replaces the optimizer with fixed 90/10 relative weights and uses only a volatility estimate — a 10-day half-life EWMA or a GARCH(1,1) model — to set the cash dilution each day. The mechanism that carries the argument is this dynamic cash dilution, which converts a fixed risky allocation into one that respects a target risk level.

What would settle it

Run a genuinely out-of-sample test: fix the 90/10 split and the 10% risk target using data only up to, say, January 2020, then backtest the rule on the remaining period with realistic transaction costs on daily rebalancing; if its Sharpe ratio no longer beats the 0.73 of the industry-only portfolio, the central claim fails.

Watch

Extended reading notes

Core claim

The paper's discovery is that a portfolio built by constrained risk allocation — specifying the fraction of total risk each asset should contribute, then scaling the whole position to a risk cap — handles crypto assets without any special machinery. The weights are found by solving one convex optimization problem, and the risk-allocation constraints reduce the search to a single scaling factor. Looking at the resulting relative weights over the sample, the authors observe that the optimizer itself keeps roughly 10% in crypto and 90% in the four industry portfolios, which motivates DD90/10, a fixed-relative-weight portfolio whose only time-varying decision is how much cash to hold. In the 2017–2024 backtest, this simple rule slightly outperforms the CRA optimizer on Sharpe ratio while keeping volatility and drawdown at the target levels, and both crypto-inclusive portfolios beat the industry-only portfolio. The conclusion is that the extreme statistical properties of crypto returns do not require abandoning standard risk-based construction.

Load-bearing premise

The fixed 90/10 split is chosen by inspecting the optimized weights over the full backtest period and then judged on that same period, so the headline Sharpe gain assumes this way of choosing the rule is not flattering the result and that ignoring trading costs would not erase the gain.

Editorial extensions

If this is right

  • A 10% crypto allocation in a risk-allocation portfolio raised the backtested Sharpe ratio from 0.73 to 1.00 without meaningfully increasing volatility or drawdown.
  • The DD90/10 rule needs no expected-return forecasts: its only inputs are a fixed relative weight and a daily volatility estimate.
  • Because the CRA problem reduces to one convex optimization plus a scalar scaling, the full method can be rebalanced every trading day at low computational cost.
  • The results are reported under a 10% annualized risk cap, a 10% cap on combined crypto weight, and daily rebalancing over September 2017 to September 2024.
  • The paper's comparison suggests that a fixed 90/10 split with cash dilution performs at least as well as the optimizer it was extracted from.

Reading between the lines

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

  • The 90/10 split was selected by looking at the full-sample CRA weights and then evaluated on that same sample; an out-of-sample test that fixes the split before the evaluation period would tell whether the simplicity is driving the result or merely being rewarded in-sample.
  • Daily rebalancing is assumed to be free; including transaction costs would erode the Sharpe advantage, particularly for crypto assets with wide spreads and in volatile periods that force large cash adjustments.
  • The specific 90/10 split and the four chosen industry portfolios are calibrated to this six-asset universe; other traditional assets or crypto tokens could yield a different best split, so the rule's sharpness as a universal prescription is untested.
  • If the mechanism is really the cash-dilution risk control rather than the exact 90/10 split, the same template should transfer to other volatile satellite assets, which is a testable extension.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. This paper considers portfolio construction that combines four traditional industry portfolios with two crypto assets (BTC, ETH). It proposes a constrained risk allocation (CRA) method that minimizes cash subject to exact risk-contribution targets, a total risk limit, and weight constraints, and shows that the nonconvex risk-allocation constraints reduce to a ray parameterized by one scalar, so the CRA problem is solved by a convex program followed by a closed-form scaling. The empirical section backtests CRA and a fixed 90/10 traditional/crypto sleeve with dynamic cash dilution (DD90/10) over September 2017 to September 2024, reporting that DD90/10 achieves annualized return 10.4%, volatility 9.8%, Sharpe 1.06, and drawdown 19.9%, versus 6.0%, 8.2%, 0.73, and 12.5% for the industry-only portfolio. The paper concludes that simple risk allocation suffices to integrate crypto assets and that a modest 10% crypto weight improves risk-adjusted performance without materially increasing risk.

Significance. If the empirical claims survive scrutiny, the paper offers a genuinely simple, implementable portfolio rule and supports the position that crypto assets do not require machine-learning-based methods. The convex optimization reduction in Section 3 is clean and correctly derived under the stated assumptions, and the paper provides a data/code repository, which are concrete strengths. The main significance is tempered by the fact that the central empirical claim rests on a single in-sample backtest with no transaction costs and with the DD90/10 rule selected from the same sample on which it is evaluated. The optimization contribution itself is solid and reproducible, but the practical 'simple and effective' conclusion needs additional empirical support.

major comments (4)
  1. [§4.1, §5.1, Table 5] The headline comparison is made before any transaction costs, although both CRA and DD90/10 are rebalanced daily and the cash-dilution weight is revised whenever the volatility estimate changes. No turnover is reported. Because crypto is the most volatile component, its weight drifts quickly, so the 90/10 sleeve will generate substantial daily trades. A conservative round-trip cost of 20–50 basis points on traded notional could plausibly reduce annualized return by more than 1–2 percentage points, which is a material share of the reported 10.4% return and could reverse the Sharpe ordering relative to the 0.73 industry-only portfolio. Please add a transaction-cost model, report turnover, or at minimum provide after-cost results for both DD90/10 and the industry-only benchmark.
  2. [§5, Figure 9, Table 5] The DD90/10 rule is reverse-engineered from the full-sample CRA holdings: Figure 9 is inspected over the entire 2017–2024 period, and the fixed 90/10 split is then evaluated on that same period in Table 5. This makes the comparable performance of DD90/10 and CRA partly an artifact of in-sample model selection. The robustness claim would be considerably strengthened by an out-of-sample evaluation, a walk-forward procedure, or at least a subperiod analysis in which the 90/10 allocation is chosen before the evaluation window.
  3. [§4.3, §6, Table 5] The conclusion states that adding a 10% crypto weight increases return and Sharpe ratio 'significantly,' but no confidence intervals, bootstrap resampling, or other uncertainty quantification is provided. With a single seven-year price path, the reported Sharpe difference (1.06 versus 0.73) may be within sampling variability, especially given the heavy tails documented in Section 1. Please add block-bootstrap intervals or a per-year/subperiod analysis to support the word 'significantly.'
  4. [§4.1, §5.1] The paper asserts parameter insensitivity and robustness to alternative volatility estimators without reporting the supporting evidence. In Section 4.1, the risk limit, maximum crypto weight, and EWMA half-lives are said to be 'chosen as reasonable values' with results 'not sensitive to these choices,' and in Section 5.1, several other volatility estimators are said to give similar results. Since these are free parameters and the DD90/10 performance is the central claim, a sensitivity table or figure is needed.
minor comments (6)
  1. [§4.3] The word 'portolios' should be 'portfolios.'
  2. [§6] The word 'contrstruction' should be 'construction.'
  3. [§4.2] In the drawdown definition, 'form' should be 'from': 'the maximum fractional drop in value from a previous high.'
  4. [§4.4] The word 'attribtutions' should be 'attributions.'
  5. [§4.1] Please clarify how the 2565 calendar days and 1729 trading days are used in the backtest, and how weekend/holiday crypto gains are incorporated when rebalancing only on trading days.
  6. [§3.2] The characterization that weights satisfying the risk-allocation constraints form a ray generated by the convex program (3) is cited to [BV24]; since this is the key mathematical step, a short proof or an explicit statement of the required positive-definiteness assumption would make the note more self-contained.

Circularity Check

1 steps flagged · score 4.0 of 10

The CRA derivation is self-contained, but the headline DD90/10 rule is selected from full-sample CRA weights and then backtested on the same sample; the 10% crypto choice also echoes the 10% constraint imposed on CRA, so the simple-rule result is partly in-sample rather than a fresh prediction.

  1. fitted input called prediction [§5 'Dynamically diluted 90/10 portfolio', Figure 9 and Table 5]
    "Figure 9 shows the relative non-cash weights, i.e., w/1T w for the combined portfolio over time. We see that, apart from 2020, the relative weights are relatively stable and evenly distributed with about 10% in crypto assets (equally split between BTC and ETH), and 90% roughly equally split between the four industry portfolios. This motivates an even simpler portfolio construction method, akin to the popular 60/40 stocks/bonds allocation."

    The DD90/10 allocation is not independently derived. The 90/10 and equal-weight choices are read off the CRA portfolio's full-sample relative weights, and the same full sample is then used to compute the DD90/10 performance in Table 5 and compare it with CRA. Moreover, the 10% crypto weight is exactly the maximum-weight constraint imposed on the CRA optimization in §4.1, so the 'discovery' that roughly 10% crypto is appropriate is in large part a restatement of the simulation input. The comparable Sharpe and return of DD90/10 is therefore an in-sample selection result, not an out-of-sample prediction. The CRA construction itself is not circular; only the simplification and its validation reduce to the same sample and the preset constraint.

full rationale

The core constrained risk allocation (CRA) derivation is self-contained: it defines a convex problem, cites standard convex optimization results for the log-barrier representation of risk-allocation weights, and solves the scaling step analytically. No equation in Section 3 is equivalent by construction to the paper's headline claim. The iterated EWMA covariance estimator is cited to the authors' own prior work [JOP+23], but it is an openly specified, parameterized estimator with code and data available, and the paper's conclusions do not rest on a uniqueness theorem imported from that paper. The main circularity burden is the DD90/10 rule. It is chosen by visual inspection of the CRA weights over the entire backtest period (Figure 9) and then evaluated on the same period (Table 5), so its comparable performance to CRA is partly an artifact of in-sample selection. Separately, the '10% crypto' split mirrors the 10% maximum crypto weight constraint already imposed on the CRA simulation, making the simple rule's central parameter an input rather than an independent finding. These issues are real but localized: the CRA framework would stand even if the DD90/10 simplification were not offered, and no definitional equation collapse or load-bearing self-citation chain is present. Hence a moderate circularity score, not a high one.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The central claim rests on a small number of hand-chosen parameters, including the risk limit, crypto cap, EWMA half-lives, and the 90/10 split, plus standard finance assumptions such as variance-based risk, long-only positions, and zero transaction costs. No new entities or mechanisms are introduced.

free parameters (7)
  • Target annualized risk limit sigma = 10% annualized (sigma = 0.1 * sqrt(D))
    Chosen by hand as a reasonable value one might use in practice (Section 4.1); the portfolio's cash dilution and asset exposure depend on it.
  • Maximum crypto weight constraint = 10% combined for BTC and ETH
    Imposed in Section 4.1; caps crypto exposure and shapes the holdings and returns of the Combined and Crypto portfolios.
  • Volatility EWMA half-life = 63 trading days
    Used in the iterated EWMA covariance estimate (Section 4.1); chosen by hand, not estimated or cross-validated.
  • Correlation EWMA half-life = 125 trading days
    Used in the iterated EWMA covariance estimate (Section 4.1); chosen by hand.
  • Portfolio risk EWMA half-life = 10 days
    Used to estimate the volatility of the unconstrained risk allocation portfolio and the 90/10 portfolio (Sections 3.2 and 5.1).
  • DD90/10 fixed allocation = 90% industries equal weight, 10% crypto equal split BTC/ETH
    Motivated by visual inspection of CRA relative weights (Section 5, Figure 9); not derived from an optimization or validated out-of-sample.
  • GARCH(1,1) estimation window = 250 trading days, refitted daily
    Used for the GARCH-based DD90/10 volatility estimator (Section 5.1); an implementation choice.
assumptions (6)
  • domain assumption Portfolio risk is adequately measured by variance w^T Sigma w, with cash treated as risk-free and non-interest-bearing.
    Used to define risk contributions and the risk limit in Section 3.1; ignores negative cash yields and tail risk beyond variance.
  • domain assumption The iterated EWMA covariance estimate with the chosen half-lives is a reliable forecast of future return covariance.
    The CRA weights and cash dilution are computed from this estimate (Section 4.1); if the forecast is biased, the ex-ante risk target is not met and the comparison is distorted.
  • domain assumption Long-only positions and no leverage, with daily rebalancing at zero transaction cost.
    The problem (2) imposes w >= 0 and a cash holding, and rebalancing is daily with no costs modeled (Section 4.1), which is optimistic for volatile crypto assets.
  • domain assumption Equal risk contribution (rho = (1/n)1) is a sensible objective without return forecasts.
    CRA is defined with a user-specified risk allocation; the paper chooses risk parity, and this choice is not derived from investor preferences (Section 3.1).
  • domain assumption The 2017-2024 sample is representative enough to support general conclusions.
    All conclusions are drawn from one historical period; no out-of-sample or cross-validation is reported (Sections 4 and 5).
  • domain assumption Heavy tails, skewness, and excess kurtosis do not invalidate covariance-based risk allocation.
    The paper notes these stylized facts but uses variance as the risk measure and the Sharpe ratio for evaluation (Sections 2.1 and 4.2).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Simple and Effective Portfolio Construction with Crypto Assets." pith.science (2026). https://pith.science/paper/CUPM6NX2

@misc{pith2026241202654,
  author       = {Pith},
  title        = {Pith review of: Simple and Effective Portfolio Construction with Crypto Assets},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CUPM6NX2}},
  note         = {Machine review of arXiv:2412.02654}
}
read the original abstract

We consider the problem of constructing a portfolio that combines traditional financial assets with crypto assets. We show that despite the documented attributes of crypto assets, such as high volatility, heavy tails, excess kurtosis, and skewness, a simple extension of traditional risk allocation provides robust solutions for integrating these emerging assets into broader investment strategies. Examination of the risk allocation holdings suggests an even simpler method, analogous to the traditional 60/40 stocks/bonds allocation, involving a fixed allocation to crypto and traditional assets, dynamically diluted with cash to achieve a target risk level.

Figures

Figures reproduced from arXiv: 2412.02654 by the authors.

Figure 1
Figure 1. Normalized prices of BTC, ETH, and SP500 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Quantile-quantile plot of log returns of BTC, ETH, and SP500. 5 [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Normalized prices of BTC, ETH, and four industry portfolios [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Estimated annualized volatilities of the six assets. Risk model. We estimate the covariance matrix of the assets using an iterated EWMA, described in detail in [JOP+23, §2.5]. We use a 63-day half-life for the volatility estimate and a 125-day half-life for the correla…
Figure 5
Figure 5. Figure 5: Estimated correlation matrices on two different dates. Return. The (realized) return of the portfolio at time t is given by w T t rt , where rt and wt are the vector of (realized) asset returns and the portfolio weights at time t, respectively. The annualized (realized…
Figure 6
Figure 6. Figure 6: Portfolio weights of the three portfolios. The cash weight is shown as uncolored. 16 [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: Portfolio values of the three portfolios. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: Annual performance metrics of the three portfolios. 18 [PITH_FULL_IMAGE:figures/full_fig_p018_8.png]
Figure 9
Figure 9. Figure 9: Relative weights of the combined portfolio. 4.4 Shapley attributions We would like to attribute the performance of the portfolio to the different asset classes. Shapley values account for each assets’s contribution to the portfolio, ensuring a fair allo￾cation. They ar…
Figure 10
Figure 10. Figure 10: Annualized volatility estimates of the 90/10 portfolio. • 90/10 portfolio. Construct a portfolio consisting of 90% equities (e.g., the four in￾dustries with equal weights) and 10% crypto (e.g., equally split between BTC and ETH). • Dynamic cash dilution. Based on an e…
Figure 11
Figure 11. Figure 11: Portfolio weights of the DD90/10 portfolios with EWMA and GARCH volatility estimators. Metric DD90/10 (EWMA) DD90/10 (GARCH) CRA Return (%) 10.4 10.1 8.2 Volatility (%) 9.8 9.7 8.2 Sharpe 1.06 1.04 1.00 Drawdown (%) 19.9 19.7 19.6 [PITH_FULL_IMAGE:figures/full_fig_p0…
Figure 12
Figure 12. Figure 12: Portfolio values of the DD90/10 and CRA portfolios. risk parity, and the other a fixed set of relative weights, with each one dynamically diluted with cash to achieve a target ex-ante risk. The addition of even a modest crypto weight of 10% increases the return and Sh…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 54 canonical work pages

  1. [1]

    Bulla and I

    J. Bulla and I. Bulla. Stylized facts of financial time series and hidden semi- M arkov models. Computational statistics & data analysis , 51(4):2192--2209, 2006

    work page 2006

  2. [2]

    Barratt and S

    S. Barratt and S. Boyd. Covariance prediction via convex optimization. Optimization and Engineering , 2022

    work page 2022

  3. [3]

    Bartram, J

    S. Bartram, J. Branke, G. De Rossi, and M. Motahari. Machine learning for active portfolio management. Journal of Financial Data Science , 3(3):9--30, 2021

    work page 2021

  4. [4]

    S. Boyd, K. Johansson, R. Kahn, P. Schiele, and T. Schmelzer. Markowitz portfolio construction at seventy. Journal of Portfolio Management , 50(8):117--160, 2024

    2024

  5. [5]

    Black and R

    F. Black and R. Litterman. Asset allocation: C ombining investor views with market equilibrium. Goldman Sachs Fixed Income Research , 115(1):7--18, 1990

    work page 1990

  6. [6]

    Black and R

    F. Black and R. Litterman. Global portfolio optimization. Financial analysts journal , 48(5):28--43, 1992

    work page 1992

  7. [7]

    F. Black. Universal hedging: Optimizing currency risk and reward in international equity portfolios. Financial Analysts Journal , 45(4):16--22, 1989

    work page 1989

  8. [8]

    Brauneis and R

    A. Brauneis and R. Mestel. Cryptocurrency-portfolios in a mean-variance framework. Finance Research Letters , 28:259--264, 2019

    work page 2019

Show all 66 references
  1. [9]

    Bollerslev

    T. Bollerslev. Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics , 31(3):307--327, 1986

    1986

  2. [10]

    Bakry, A

    W. Bakry, A. Rashid, S. Al-Mohamad, and N. El-Kanj. Bitcoin and portfolio diversification: A portfolio optimization approach. Journal of Risk and Financial Management , 14(7):282, 2021

    2021

  3. [11]

    Burggraf

    T. Burggraf. Risk-based portfolio optimization in the cryptocurrency world. Available at SSRN 3454764 , 2019

    2019

  4. [12]

    Boyd and L

    S. Boyd and L. Vandenberghe. Convex Optimization . Cambridge University Press, 2004

    2004

  5. [13]

    Boyd and L

    S. Boyd and L. Vandenberghe. Convex optimization additional exercises. https://github.com/cvxgrp/cvxbook_additional_exercises, 2024

    2024

  6. [14]

    Choueifaty and Y

    Y. Choueifaty and Y. Coignard. Toward maximum diversification. The Journal of Portfolio Management , 35(1):40--51, 2008

    2008

  7. [15]

    Ciolli and J

    J. Ciolli and J. Kara. Paul K rugman: Bitcoin is a more obvious bubble than housing was. Business Insider , 2013

    2013

  8. [16]

    Chakrabarty and A

    S. Chakrabarty and A. Kanaujiya. Mathematical Portfolio Theory and Analysis . Springer Nature, 2023

    2023

  9. [17]

    u t \"u nc \

    G. Cornuejols and R. T \"u t \"u nc \"u . Optimization methods in finance , volume 5. Cambridge University Press, 2006

    2006

  10. [18]

    Diamond and S

    S. Diamond and S. Boyd. CVXPY : A P ython-embedded modeling language for convex optimization. Journal of Machine Learning Research , 17(83):1--5, 2016

    2016

  11. [19]

    Engle and T

    R. Engle and T. Bollerslev. Modelling the persistence of conditional variances. Econometric reviews , 5(1):1--50, 1986

    1986

  12. [20]

    R. Engle. Autoregressive conditional heteroscedasticity with estimates of the variance of U nited K ingdom inflation. Econometrica: J ournal of the econometric society , pages 987--1007, 1982

    1982

  13. [21]

    R. Engle. Dynamic conditional correlation. Journal of Business & Economic Statistics , 20(3):339--350, 2002

    2002

  14. [22]

    Feng and D

    Y. Feng and D. Palomar. Scrip: Successive convex optimization methods for risk parity portfolio design. IEEE Transactions on Signal Processing , 63(19):5285--5300, 2015

    2015

  15. [23]

    K. French. Kenneth french data library. https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html\#Research, 2024

    2024

  16. [24]

    Fryer, I

    D. Fryer, I. Str \"u mke, and H. Nguyen. Shapley values for feature selection: T he good, the bad, and the axioms. IEEE Access , 9:144352--144360, 2021

    2021

  17. [25]

    Gunjan and S

    A. Gunjan and S. Bhattacharyya. A brief review of portfolio optimization techniques. Artificial Intelligence Review , 56(5):3847--3886, 2023

    2023

  18. [26]

    Ghosh, E

    B. Ghosh, E. Bouri, J. Bum Wee, and N. Zulfiqar. Return and volatility properties: Stylized facts from the universe of cryptocurrencies and NFTs . Research in International Business and Finance , 65:101945, 2023

    2023

  19. [27]

    R. Geary. Moments of the ratio of the mean deviation to the standard deviation for normal samples. Biometrika , pages 295--307, 1936

    1936

  20. [28]

    R. Geary. Testing for normality. Biometrika , 34(3/4):209--242, 1947

    1947

  21. [29]

    Grinold and R

    R. Grinold and R. Kahn. Active portfolio management . McGraw-Hill, 2000

    2000

  22. [30]

    Giudici, P

    P. Giudici, P. Pagnottoni, and G. Polinesi. Network models to enhance automated cryptocurrency portfolio management. Frontiers in artificial intelligence , 3:22, 2020

    2020

  23. [31]

    I. Gurrib. Machine learning and portfolio management: A review. Annals of Mathematics and Computer Science , 5:31--43, 2022

    2022

  24. [32]

    Holovatiuk

    O. Holovatiuk. Cryptocurrencies as an asset class in portfolio optimisation. Central European Economic Journal , 7(54):33--55, 2020

    2020

  25. [33]

    A. Hu, C. Parlour, and U. Rajan. Cryptocurrencies: Stylized facts on a new investible instrument. Financial Management , 48(4):1049--1068, 2019

    2019

  26. [34]

    Y. Hu, S. Rachev, and F. Fabozzi. Modelling crypto asset price dynamics, optimal crypto portfolio, and crypto option valuation. arXiv preprint arXiv:1908.05419 , 2019

    1908 arXiv

  27. [35]

    Huettner and M

    F. Huettner and M. Sunder. Axiomatic arguments for decomposing goodness of fit according to S hapley and O wen values. 2012

    2012

  28. [36]

    Jiang and J

    Z. Jiang and J. Liang. Cryptocurrency portfolio management with deep reinforcement learning. In 2017 Intelligent systems conference (IntelliSys) , pages 905--913. IEEE, 2017

    2017

  29. [37]

    Johansson, G

    K. Johansson, G. Ogut, M. Pelger, T. Schmelzer, and S. Boyd. A simple method for predicting covariance matrices of financial returns. Foundations and Trends in Econometrics , 12(4):324--407, 2023

    2023

  30. [38]

    Jiang, K

    L. Jiang, K. Wu, G. Zhou, and Y. Zhu. Stock return asymmetry: Beyond skewness. Journal of Financial and Quantitative Analysis , 55(2):357--386, 2020

    2020

  31. [39]

    Jiang, D

    Z. Jiang, D. Xu, and J. Liang. A deep reinforcement learning framework for the financial portfolio management problem. arXiv preprint arXiv:1706.10059 , 2017

    2017 arXiv

  32. [40]

    Karim, Mohamed E

    M. Karim, Mohamed E. Rasid, A. Noman, and L. Yarovaya. Exploring asymmetries in cryptocurrency intraday returns and implied volatility: New evidence for high-frequency traders. International Review of Financial Analysis , page 103617, 2024

    2024

  33. [41]

    P. Krugman. Golden cyberfetters. The New York Times , 2011

    2011

  34. [42]

    P. Krugman. Adam S mith hates bitcoin. Business Insider , 2013

    2013

  35. [43]

    u t \"u nc \

    P. Kolm, R. T \"u t \"u nc \"u , and F. Fabozzi. 60 years of portfolio optimization: Practical challenges and current trends. European Journal of Operational Research , 234(2):356--371, 2014

    2014

  36. [44]

    Kelly and D

    B. Kelly and D. Xiu. Financial machine learning. Foundations and Trends in Finance , 13(3-4):205--363, 2023

    2023

  37. [45]

    Lucarelli and M

    G. Lucarelli and M. Borrotti. A deep Q -learning portfolio management framework for the cryptocurrency market. Neural Computing and Applications , 32:17229--17244, 2020

    2020

  38. [46]

    Liu and A

    Y. Liu and A. Tsyvinski. Risks and returns of cryptocurrency. The Review of Financial Studies , 34(6):2689--2727, 2021

    2021

  39. [47]

    Markowitz

    H. Markowitz. Portfolio selection. Journal of Finance , 7(1):77--91, 1952

    1952

  40. [48]

    Moehle, S

    N. Moehle, S. Boyd, and A. Ang. Portfolio performance attribution via shapley value. arXiv preprint arXiv:2102.05799 , 2021

    2021 arXiv

  41. [49]

    A. Meucci. Fully flexible views: Theory and practice. arXiv preprint arXiv:1012.2848 , 2010

    2010 arXiv

  42. [50]

    R. Michaud. The M arkowitz optimization enigma: I s `optimized' optimal? Financial analysts journal , 45(1):31--42, 1989

    1989

  43. [51]

    Malmsten and T

    H. Malmsten and T. Ter \"a svirta. Stylized facts of financial time series and three popular models of volatility. European Journal of pure and applied mathematics , 3(3):443--477, 2010

    2010

  44. [52]

    Nakamoto

    S. Nakamoto. A peer-to-peer electronic cash system. Bitcoin.--URL: https://bitcoin. org/bitcoin. pdf , 4(2):15, 2008

    2008

  45. [53]

    R. Narang. Inside the Black Box: A Simple Guide to Systematic Investing . J. Wiley & Sons, 2024

    2024

  46. [54]

    Owen and C

    A. Owen and C. Prieur. On S hapley value for measuring importance of dependent inputs. SIAM/ASA Journal on Uncertainty Quantification , 5(1):986--1002, 2017

    2017

  47. [55]

    D. Palomar. Portfolio Optimization . Cambridge University Press, 2024

    2024

  48. [56]

    Platanakis and A

    E. Platanakis and A. Urquhart. Portfolio management with cryptocurrencies: The role of estimation risk. Economics Letters , 177:76--80, 2019

    2019

  49. [57]

    E. Qian. Risk parity and diversification. The Journal of Investing , 20(1):119--127, 2011

    2011

  50. [58]

    Ramkumar

    G. Ramkumar. Cryptocurrency portfolio construction using machine learning models. In Contemporary Trends and Challenges in Finance: Proceedings from the 6th Wroclaw International Conference in Finance , pages 103--122. Springer, 2021

    2021

  51. [59]

    Ruiz Roque da Silva, E

    I. Ruiz Roque da Silva, E. Junior, and P. Balbi. Cryptocurrencies trading algorithms: A review. Journal of Forecasting , 41(8):1661--1668, 2022

    2022

  52. [60]

    Rockafellar and S

    T. Rockafellar and S. Uryasev. Optimization of conditional value-at-risk. Journal of Risk , 2:21--42, 2000

    2000

  53. [61]

    L. Shapley. A value for n -person games. Contribution to the Theory of Games , 2, 1953

    1953

  54. [62]

    Sheppard

    K. Sheppard. Arch documentation. 2024

    2024

  55. [63]

    L. Wu, Y. Feng, and D. Palomar. General sparse risk parity portfolio design via successive convex optimization. Signal Processing , 170:107433, 2020

    2020

  56. [64]

    J. Yarow. Paul K rugman responds to all the people throwing around his old internet quote. Business Insider , 2013

    2013

  57. [65]

    D. Yermack. Is bitcoin a real currency? A n economic appraisal. In Handbook of digital currency , pages 29--40. Elsevier, 2024

    2024

  58. [66]

    Zhang, H

    H. Zhang, H. Singh, M. Ghassemi, and S. Joshi. W hy did the model fail? A ttributing model performance changes to distribution shifts. 2023

    2023

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.