Pith. sign in

REVIEW 4 major objections 5 minor 70 references

Fragility of Minimum-Variance Portfolios

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Minimum-variance portfolio instability is a threshold effect, not just estimation noise

desk verdict Mathematically sound and useful closed forms, but the headline mechanism is not cleanly isolated from conditioning effects and some real-data claims overreach; worth refereeing with revisions. read the letter →

arxiv 2607.18624 v1 pith:HYHRGHBH submitted 2026-07-21 math.OC

Daniel Ovalle , Carl D. Laird , Ignacio E. Grossmann , Javier Peña This is my paper · ORCID
classification math.OC MSC 91G1090C2090C25
keywords minimum-varianceportfoliofragilitythresholdeffectsblock-diagonalcorrelationfixed-pointshrinkageactivesetclustering
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 seeks to explain why long-only minimum-variance portfolios are so sensitive to covariance estimation error. It shows that under constant or block-constant correlation, the optimal portfolio has a closed form in which assets whose inverse volatilities fall below a fixed-point threshold are dropped. Fragility occurs when that threshold sits close to an asset's inverse volatility, so a tiny perturbation in correlation or volatility flips the asset in or out of the portfolio. This mechanism is explicit and testable, and it motivates shrinkage corrections that move the threshold away from breakpoints rather than shrinking the whole covariance. If correct, the paper converts a vague 'estimation error' story into a precise, fixable cause.

What carries the argument

The fixed-point threshold θ̄ (and its block analog θ̄_i + θ̂) is the central object. It is the cutoff in inverse-volatility space determined by correlation ρ and the positive-part truncation; because it solves θ̄ = ρ/(1−ρ) Σ (θ_j − θ̄)_+, it encodes the whole coupling between correlation and volatilities. The paper's proofs rely on the reparametrization x = Θy/(1^T Θy) and the KKT conditions, reducing the portfolio problem to a convex quadratic program whose solution is the truncation. This object carries the argument because proximity of θ̄ to a breakpoint is defined as the exact fragility condition.

What would settle it

Construct an exact block-constant covariance with ordered inverse volatilities arranged so that θ̄ is far from every breakpoint (e.g., satisfying θ_{i+1} << θ̄ << θ_i); then perturb ρ and all σ_i by small amounts and check that the active set and weights remain continuous. If the active set changes despite this separation, the threshold mechanism is wrong. Conversely, take the paper's fragile three-asset example (σ=(1,1.8,2), ρ=0.48) and deliberately nudge σ3; the active set should flip exactly when θ̄ crosses θ_3.

Watch

Extended reading notes

Core claim

Under correlation matrices that are constant within blocks, the long-only minimum-variance portfolio equals a normalized truncation of inverse volatilities: y ∝ (θ − θ̄1)_+, where θ̄ solves a fixed-point equation. The authors prove that the active set of assets is determined by which inverse volatilities exceed θ̄, and that the portfolio is fragile exactly when θ̄ is near one of the breakpoints θ_i or θ_{i+1}: small perturbations in ρ, σ_i, or σ_{i+1} can shift the active set and cause discontinuous weight changes. They extend this to block-diagonal matrices, where each block has its own threshold θ̄_i plus a common global threshold θ̂, coupled through a fixed-point system. They then propose

Load-bearing premise

The analysis assumes the true correlation structure is approximately block-constant with known blocks and that the clustering procedure recovers those blocks; if the fitted blocks misplace the breakpoints, the threshold corrections are aimed at the wrong boundaries and the mechanism no longer predicts the real portfolio.

Editorial extensions

If this is right

  • If θ̄ stays a safe distance away from every inverse-volatility breakpoint, the active set is locally invariant and small estimation errors cannot flip assets in or out.
  • The closed form gives a diagnostic: practitioners can compute θ̄ from an estimated covariance and flag portfolios where θ̄ is close to a breakpoint.
  • Correlation-aware shrinkage that projects θ̄ away from breakpoints keeps the portfolio sparse and close to the minimum-variance solution, unlike global shrinkage which densifies it.
  • The block-diagonal fixed-point system shows fragility can be localized: one block operating near its threshold is enough to destabilize the whole allocation.
  • Because the block structure can be recovered by clustering, the mechanism extends to general covariance matrices without requiring an exogenous grouping.

Reading between the lines

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

  • A natural test is to monitor the distance min_i |θ̄ − θ_i| over rolling windows; if the paper is right, turnover and realized variance should spike when this distance is small, providing an early-warning statistic.
  • The same threshold mechanism may translate to mean-variance problems, where expected-return estimates enter as an additional drift term; the paper leaves this open.
  • The largest-gap clustering cutoff is only one way to recover blocks; the fragility diagnosis might be sensitive to clustering method, and comparing alternative linkages could reveal whether the mechanism is robust to block mis-specification.
  • The correction parameter ε in the projection heuristic is chosen as max θ_i / n; a data-driven calibration of ε could improve the trade-off further, though the paper does not explore it.
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.

Referee Report

4 major / 5 minor

Summary. The paper derives closed-form solutions for long-only minimum-variance portfolios under constant and block-constant correlation structures (Propositions 1 and 2), showing that the optimal portfolio is determined by a fixed-point threshold θ̄ (and, in the block case, coupled thresholds θ̂+θ̄_i) that truncates inverse volatilities. Fragility is attributed to proximity of these thresholds to the ordered θ_i breakpoints. Motivated by this insight, the authors propose structured shrinkage methods (BDS-Obliv, BDS-Aware, BDS-Direct) and evaluate them against classical and hierarchical benchmarks in controlled simulations and on Fama-French industry returns. Proposition 3 gives the least-squares fit of the block model.

Significance. The closed-form results are a genuine contribution: they make the support-selection mechanism of the long-only minimum-variance problem explicit under a structured correlation model, and the derivations are self-contained and algebraically sound. The BDS family is clearly described, an open-source repository is promised, and the benchmark comparison is unusually broad. However, the central mechanism—threshold distance as the driver of fragility—is not directly tested; the real-data 'confirmation' is in-sample, and the language of discontinuous weight changes is technically inaccurate. If a controlled distance sweep is added and confirms the mechanism, the paper would provide a useful structural explanation and a well-motivated shrinkage procedure.

major comments (4)
  1. [§3.1, Eq. (8)–(11); Examples 1–3] The paper's central claim—fragility is governed by the distance of θ̄ to the breakpoints θ_i—is asserted but never tested by varying that distance. All three examples place parameters at or near the boundary (Ex. 1: ρ=σ1/σ2 with σ1≈σ2, so the covariance is near-singular; Ex. 2: θ3≈θ̄; Ex. 3: θ_i≈θ̂+θ̄_i with all θ_i equal). These constructions cannot separate threshold proximity from generic ill-conditioning or block degeneracy. The authors should sweep δ = min_i |θ̄−θ_i| while holding n, ρ, volatility spread, estimator, and evaluation noise fixed, and show that Area, turnover, and active-set switch frequency decrease with δ. Without this control, the robustification rationale in §5 is not empirically established.
  2. [§4.2, Fig. 6 and §5.3.3] The real-data 'sharp prediction' is in-sample and circular. The block model, thresholds θ̂+θ̄_i, and asset-to-threshold distances are estimated from the full 2018–2025 residualized sample, and the same sample is then used to observe that Mastercard and the Energy names toggle. The 'True' variance in Fig. 6 also uses the full-sample residual covariance as the oracle, making Area an in-sample quantity. This is a descriptive fit, not an external confirmation of the mechanism. Please re-estimate the block structure and thresholds on rolling or training windows and test the predicted toggling out-of-sample, or at least explicitly label the claim as in-sample.
  3. [§3.1, text after Eq. (10)] The statement that an active-set shift 'changes which components are truncated ... leading to a discontinuous change in the portfolio weights (3)' is inaccurate. In the positive-part solution (8), an asset at the boundary has (θ_i−θ̄)_+=0, so y_i=0 and x_i=0; the mapping from parameters to weights is continuous. The fragility is a kink/derivative discontinuity, not a jump in x. The 'abrupt switches' in the rolling experiments come from estimated parameters jumping across the boundary across rebalancing dates. Please correct this characterization and, if possible, quantify the local Lipschitz constant or derivative of x near the threshold.
  4. [§5.3 and §6.1] The block approximation (12) is assumed to be a faithful representation of general covariance matrices, and the clustering pipeline is validated only indirectly. In §6.1 the DGP is block-diagonal by construction, so block recovery is favorable; in the factor model and real data there is no ground-truth partition, so the inferred blocks are not checked. The generality of the BDS results depends on the largest-gap/single-linkage rule (19)–(22) recovering meaningful structure. Please report cluster-recovery metrics where true blocks are known (e.g., adjusted Rand index) and a sensitivity analysis of BDS performance to the dendrogram cutoff τ.
minor comments (5)
  1. [§4.2, Example 3] Notation: 'ˆθ1+θ̄' should be 'θ̂+θ̄_i'; there is no θ̂_1 defined in the paper.
  2. [§5.3.2, Fig. 9 caption] The caption says the threshold is τ = ½(ℓ_k*+ℓ_k*+1)+ε, but Eq. (22) defines τ without the +ε term. Please reconcile.
  3. [Figures 4, 7] The axis labels use 'Ast.'; this should be 'Asset' throughout the figure panels.
  4. [§5.1.2, Eq. (15)] The ε heuristic is effectively a free parameter; no sensitivity analysis with respect to ε is provided. Either report sensitivity or state that results are invariant over a reasonable range.
  5. [§4.2, Fig. 6] Please define precisely what covariance is used as 'true' in the real-data Area computation; the text implies the full-sample residual covariance, which should be stated explicitly.

Circularity Check

1 steps flagged · score 4.0 of 10

In-sample 'prediction' in §5.3.3 is fitted to the same data it claims to confirm; the closed-form derivation itself is independent.

  1. fitted input called prediction [Section 5.3.3, final paragraph (after Figure 10)]
    "The block fixed-point system therefore makes a sharp prediction: assets operating at or near their block thresholds (Mastercard and the Energy names) should be the most susceptible to toggling as the estimated covariance fluctuates across rolling windows, while assets with coordinates safely above their breakpoints should remain persistently active. Figure 6 confirms this prediction closely..."

    The thresholds θ̂+θ̄_i used for the prediction are obtained from the block parameters ρ̂, ρ̂_i estimated by Proposition 3 via least-squares projection onto the same full-sample residual correlation matrix that also generates the rolling-window portfolios in Figure 6. The same data set therefore both determines which assets are near the thresholds and supplies the toggling behavior offered as confirmation. The 'sharp prediction' is an in-sample restatement of the fitted model's implications, not an out-of-sample test, so the claimed confirmation is descriptive rather than independent evidence.

full rationale

The analytical core of the paper is self-contained and not circular. Lemma 1 is an exact change of variables; Proposition 1 and Proposition 2 are derived from KKT conditions and algebraic manipulation of the positive-part operator, with no fitted parameters or self-citations. The central fragility claim in Section 3.1 follows directly from formula (8): if the fixed-point threshold θ̄ sits near a breakpoint θ_i or θ_{i+1}, small parameter perturbations can change the active set. That is a mathematical property of the derived solution, not an input. The block-diagonal extension in Proposition 2 is likewise a direct derivation, and Proposition 3 is a straightforward least-squares projection. There is no uniqueness theorem imported from the authors' prior work, and no ansatz smuggled in via self-citation. The only step that approaches a circular pattern is the real-data 'confirmation' in Section 5.3.3: the block structure and thresholds are estimated from the full-sample residual covariance, and the same sample is used both to identify assets near thresholds and to observe which assets toggle in the rolling-window backtest. This is an in-sample description rather than an external validation, so the paper overstates the evidential value of Figure 6. It does not, however, undermine the mathematical derivation or the controlled simulation comparisons, which are independent of this example. Because the central claim retains independent content and the circularity is confined to a supporting empirical illustration, the score is moderate rather than high.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The central theorems need only convex QP KKT and positive definiteness; those are standard and proved in appendices. The economically load-bearing inputs are the block-constant correlation model, the clustering rule, and the margin ε, the last being the only hand-set free parameter. No new particles, mediators, or externally testable entities are introduced.

free parameters (1)
  • epsilon (robustness margin) = max_i θ_i / n (Eq. 15)
    The robustness margin ε in Section 5.1.2 controls projection of θ̄ onto the robust-feasible set F. It is a hand-set tuning knob with no link to the actual estimation error or to the data; the paper suggests a heuristic default but does not derive it from first principles.
assumptions (4)
  • domain assumption The covariance dependence structure is adequately captured by a block-diagonal constant-correlation matrix Ω=(1−ρ)Diag(Ω_i)+ρ11^T, or its single-block special case.
    Invoked throughout Sections 3–5; it is the premise on which the closed-form expressions and threshold mechanism rest. Section 4.1 introduces the block form as a stylized approximation of clustered markets.
  • domain assumption Single-linkage hierarchical clustering with correlation distance and the largest-gap cutoff (22) recover the economically meaningful blocks without an exogenous number of clusters.
    Section 5.3.1–5.3.2; the BDS pipeline depends on this clustering recovering dependence groups from noisy empirical correlations. No theoretical guarantee is provided that the recovered partition preserves the threshold geometry of the true covariance.
  • domain assumption Residualizing returns with respect to the market factor leaves a correlation matrix whose dominant structure is block-like, and the sample covariance estimated on the full window can be treated as true V for computing Area.
    Section 4.2 and Figure 6; this makes the subsequent 'prediction' in Section 5.3.3 an in-sample description. If V is not the true covariance, the Area metric is not an oracle-based loss.
  • standard math KKT conditions are necessary and sufficient for the convex QPs (1) and (4), and covariance matrices are positive semidefinite as specified.
    Used in Lemma 1 and Appendices A.1–A.4 to derive the closed-form solutions and fixed-point equations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fragility of Minimum-Variance Portfolios." pith.science (2026). https://pith.science/paper/HYHRGHBH

@misc{pith2026260718624,
  author       = {Pith},
  title        = {Pith review of: Fragility of Minimum-Variance Portfolios},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HYHRGHBH}},
  note         = {Machine review of arXiv:2607.18624}
}
read the original abstract

Minimum-variance portfolios are well known to be highly sensitive to covariance estimation error. In this paper, we show that by imposing a block diagonal correlation structure, we can derive closed-form expressions for long-only minimum-variance portfolios that make this fragility explicit. These analytical solutions reveal that fragility is driven by threshold effects arising from the interaction between correlation structure and the assets' volatilities. Motivated by the latter insight, we propose robustification approaches that can be interpreted as structured shrinkage schemes that selectively attenuate unstable coupling while preserving the dominant risk structure. Unlike global shrinkage techniques, the proposed corrections are analytically grounded, require minimal tuning, and remain closely aligned with the minimum-variance solution. The framework extends naturally to general covariance matrices through clustering-based approximations. Empirical results on controlled simulations highlight a clear regime dependence. In homogeneous volatility settings, correlation-oblivious shrinkage achieves the best trade-off between risk and stability. In contrast, under heterogeneous volatility, correlation-aware shrinkage performs best by inducing sparsity and avoiding exposure to high-risk assets. Across regimes, the proposed methods consistently reduce out-of-sample variance and turnover relative to classical and clustering-based benchmarks, providing a principled and practical approach to robust portfolio construction.

Figures

Figures reproduced from arXiv: 2607.18624 by the authors.

Figure 1
Figure 1. Fragile long-only Markowitz portfolio with [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Fragile three-asset long-only Markowitz portfolio with [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. , while the resulting variance dynamics and portfolio weights are reported in [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figures from the paper (24 more)
Figure 4
Figure 4. Figure 4: Fragility long-only portfolio under the three-block structure. [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]
Figure 5
Figure 5. Figure 5: Residual correlation matrix Ω after removing the market factor from monthly returns (2018–2024). Sectoral clustering becomes visible once the common market component is eliminated. 0 50 100 150 200 250 Day 0.0100 0.0105 0.0110 0.0115 Variance Long-Only Markowitz | Area…
Figure 6
Figure 6. Figure 6: Long-only minimum-variance portfolio for the residualized real-data universe. [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: Eight-asset universe under two volatility configurations. Top two panels: uniform volatilities [PITH_FULL_IMAGE:figures/full_fig_p021_7.png]
Figure 8
Figure 8. Figure 8: Residual correlation matrix Ω of the seven-ticker market example of Section 4.2 (monthly returns, 2018-2025, market factor removed). We show the assets in alphabetical order; no block structure is visible along with the same matrix reordered by the single-linkage parti…
Figure 9
Figure 9. Figure 9: Largest-gap cut applied to the single-linkage dendrogram of [PITH_FULL_IMAGE:figures/full_fig_p025_9.png]
Figure 10
Figure 10. Figure 10: Residual correlation matrix Ω after removing the market factor from monthly returns (discussed in Section 4.2 [PITH_FULL_IMAGE:figures/full_fig_p027_10.png]
Figure 11
Figure 11. Figure 11: Covariance matrices generated by the block-diagonal data-generating process at [PITH_FULL_IMAGE:figures/full_fig_p030_11.png]
Figure 12
Figure 12. Figure 12: Area–turnover Pareto scatter pooled over [PITH_FULL_IMAGE:figures/full_fig_p031_12.png]
Figure 13
Figure 13. Figure 13: log10(areamethod/areaLOM) against universe size n ∈ {50, . . . , 250} at w = 2n. LOM serves as the strongest non-BDS baseline. Shaded bands show 95% confidence intervals over 10 seeds. The annotation at n = 250 reports the fold-gap in area between BDS-Direct and HRP. …
Figure 14
Figure 14. Figure 14: Correlation matrix of a Lopez de Prado [ [PITH_FULL_IMAGE:figures/full_fig_p034_14.png]
Figure 15
Figure 15. Figure 15: Area–turnover Pareto scatter pooled over all six Lopez de Prado configurations and 10 seeds. [PITH_FULL_IMAGE:figures/full_fig_p035_15.png]
Figure 16
Figure 16. Figure 16: Realized annualized volatility vs. mean monthly turnover for 26 portfolio methods on Fama-French [PITH_FULL_IMAGE:figures/full_fig_p038_16.png]
Figure 17
Figure 17. Figure 17: 24-month rolling realized volatility for selected methods on Fama-French 49 industry XSMOM [PITH_FULL_IMAGE:figures/full_fig_p038_17.png]
Figure 18
Figure 18. Figure 18: Pareto scatter at n = 100 and σs = 1.2 under increasing cross-block correlation. BDS-Aware remains on or near the frontier across all ρ. BDS-Obliv degrades as cross-block dependence increases, while BDS-Direct becomes the lowest-area method at higher ρ. Covariance est…
Figure 19
Figure 19. Figure 19: Pareto scatter by covariance-estimator family at [PITH_FULL_IMAGE:figures/full_fig_p051_19.png]
Figure 20
Figure 20. Figure 20: Area–turnover Pareto frontier across factor-loading noise levels in the Lopez de Prado model [PITH_FULL_IMAGE:figures/full_fig_p052_20.png]
Figure 21
Figure 21. Figure 21: Method ranks across (n,regime) cells in the calibrated sweep (w = 2n, σs = 1.2). Each cell shows area rank (lower-left) and turnover rank (upper-right), where 1 is best (green). BDS-Direct and BDS-Aware dominate area in heterogeneous regimes, while BDS-Obliv dominates…
Figure 22
Figure 22. Figure 22: Pareto frontier by universe size n in the heterogeneous regime (w = 2n). BDS methods define the frontier at every n, with a widening performance gap as n increases. 65 [PITH_FULL_IMAGE:figures/full_fig_p065_22.png]
Figure 23
Figure 23. Figure 23: Area ranks across (n,regime) for w ∈ {2n, 3n}. Each cell shows w = 2n (lower-left) and w = 3n (upper-right), with 1 as best (green). Ordering is fixed to w = 2n. Rankings are stable across window lengths. 66 [PITH_FULL_IMAGE:figures/full_fig_p066_23.png]
Figure 24
Figure 24. Figure 24: Change in realized volatility relative to sample covariance baseline in the heterogeneous regime [PITH_FULL_IMAGE:figures/full_fig_p067_24.png]
Figure 25
Figure 25. Figure 25: Pareto frontiers for six Lopez configurations (10 seeds). BDS methods define the frontier in all [PITH_FULL_IMAGE:figures/full_fig_p068_25.png]
Figure 26
Figure 26. Figure 26: Realized annualized volatility vs. mean monthly turnover for each of 26 portfolio methods on [PITH_FULL_IMAGE:figures/full_fig_p069_26.png]
Figure 27
Figure 27. Figure 27: 24-month rolling realized annualized volatility (%) for nine portfolio methods on the Fama [PITH_FULL_IMAGE:figures/full_fig_p069_27.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

70 extracted references · 2 linked inside Pith

  1. [1]

    Portfolio selection.Journal of Finance, 7(1):77–91, 1952

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

    1952

  2. [2]

    The markowitz optimization enigma: Is ‘optimized’optimal?Financial analysts journal, 45(1):31–42, 1989

    Richard O Michaud. The markowitz optimization enigma: Is ‘optimized’optimal?Financial analysts journal, 45(1):31–42, 1989

    1989

  3. [3]

    On the sensitivity of mean-variance-efficient portfolios to changes in asset means: some analytical and computational results.The review of financial studies, 4(2):315–342, 1991

    Michael J Best and Robert R Grauer. On the sensitivity of mean-variance-efficient portfolios to changes in asset means: some analytical and computational results.The review of financial studies, 4(2):315–342, 1991

    1991

  4. [4]

    The effect of errors in means, variances, and covariances on optimal portfolio choice.Journal of Portfolio Management, 19(2):6–11, 1993

    Vijay K Chopra, William T Ziemba, et al. The effect of errors in means, variances, and covariances on optimal portfolio choice.Journal of Portfolio Management, 19(2):6–11, 1993

    1993

  5. [5]

    Cambridge University Press Cambridge, UK, 1993

    Peter Muller.Empirical tests of biases in equity portfolio optimization. Cambridge University Press Cambridge, UK, 1993

    1993

  6. [6]

    Efficient asset management: a practical guide to stock portfolio optimization and asset allocation., 2001

    Richard O Michaud and Tongshu Ma. Efficient asset management: a practical guide to stock portfolio optimization and asset allocation., 2001

    2001

  7. [7]

    Portfolio choice problems

    Michael W Brandt. Portfolio choice problems. InHandbook of financial econometrics: Tools and techniques, pages 269–336. Elsevier, 2010

    2010

  8. [8]

    Efficient portfolio selection in a large market.Journal of Financial Econometrics, 14(3):496–524, 2016

    Jiaqin Chen and Ming Yuan. Efficient portfolio selection in a large market.Journal of Financial Econometrics, 14(3):496–524, 2016

    2016

Show all 70 references
  1. [9]

    Minimum-variance portfolios in the us equity market.Journal of Portfolio Management, 33(1):10, 2006

    Roger Clarke, Harindra De Silva, and Steven Thorley. Minimum-variance portfolios in the us equity market.Journal of Portfolio Management, 33(1):10, 2006

    2006

  2. [10]

    Optimal versus naive diversification: How inefficient is the 1/n portfolio strategy?The review of Financial studies, 22(5):1915–1953, 2009

    Victor DeMiguel, Lorenzo Garlappi, and Raman Uppal. Optimal versus naive diversification: How inefficient is the 1/n portfolio strategy?The review of Financial studies, 22(5):1915–1953, 2009

    1915

  3. [11]

    Balanced baskets: A new approach to trading and hedging risks.Journal of Investment Strategies (Risk Journals), 1(4), 2012

    David H Bailey and Marcos Lopez de Prado. Balanced baskets: A new approach to trading and hedging risks.Journal of Investment Strategies (Risk Journals), 1(4), 2012

    2012

  4. [12]

    Building diversified portfolios that outperform out-of-sample.Journal of Portfolio Management, 2016

    Marcos Lopez de Prado. Building diversified portfolios that outperform out-of-sample.Journal of Portfolio Management, 2016

    2016

  5. [13]

    CRC press, 2013

    Thierry Roncalli.Introduction to risk parity and budgeting. CRC press, 2013. 55

    2013

  6. [14]

    Minimum-variance portfolio composition.Journal of Portfolio Management, 37(2):31, 2011

    Roger Clarke, Harindra De Silva, and Steven Thorley. Minimum-variance portfolio composition.Journal of Portfolio Management, 37(2):31, 2011

    2011

  7. [15]

    Enhanced portfolio optimization.Financial Analysts Journal, 77(2):124–151, 2021

    Lasse Heje Pedersen, Abhilash Babu, and Ari Levine. Enhanced portfolio optimization.Financial Analysts Journal, 77(2):124–151, 2021

    2021

  8. [16]

    Spectrum estimation for large dimensional covariance matrices using random matrix theory.The Annals of Statistics, pages 2757–2790, 2008

    Noureddine El Karoui. Spectrum estimation for large dimensional covariance matrices using random matrix theory.The Annals of Statistics, pages 2757–2790, 2008

    2008

  9. [17]

    Cleaning large correlation matrices: tools from random matrix theory.Physics Reports, 666:1–109, 2017

    Joël Bun, Jean-Philippe Bouchaud, and Marc Potters. Cleaning large correlation matrices: tools from random matrix theory.Physics Reports, 666:1–109, 2017

    2017

  10. [18]

    Geometric representation of high dimension, low sample size data.Journal of the Royal Statistical Society Series B: Statistical Methodology, 67(3): 427–444, 2005

    Peter Hall, James Stephen Marron, and Amnon Neeman. Geometric representation of high dimension, low sample size data.Journal of the Royal Statistical Society Series B: Statistical Methodology, 67(3): 427–444, 2005

    2005

  11. [19]

    Thestatisticsandmathematicsofhighdimension low sample size asymptotics.Statistica Sinica, 26(4):1747, 2016

    DanShen, HaipengShen, HongtuZhu, andJSMarron. Thestatisticsandmathematicsofhighdimension low sample size asymptotics.Statistica Sinica, 26(4):1747, 2016

    2016

  12. [20]

    A survey of high dimension low sample size asymptotics.Australian & New Zealand journal of statistics, 60(1):4–19, 2018

    Makoto Aoshima, Dan Shen, Haipeng Shen, Kazuyoshi Yata, Yi-Hui Zhou, and James S Marron. A survey of high dimension low sample size asymptotics.Australian & New Zealand journal of statistics, 60(1):4–19, 2018

    2018

  13. [21]

    High-dimensionality effects in the markowitz problem and other quadratic programs with linear constraints: Risk underestimation.Ann

    Noureddine El Karoui. High-dimensionality effects in the markowitz problem and other quadratic programs with linear constraints: Risk underestimation.Ann. Statist., 38(1):3487–3566, 2010

    2010

  14. [22]

    On the realized risk of high-dimensional markowitz portfolios.SIAM Journal on Financial Mathematics, 4(1):737–783, 2013

    Noureddine El Karoui. On the realized risk of high-dimensional markowitz portfolios.SIAM Journal on Financial Mathematics, 4(1):737–783, 2013

    2013

  15. [23]

    Tractable stochastic analysis in high dimensions via robust optimization.Mathematical programming, 134(1):23–70, 2012

    Chaithanya Bandi and Dimitris Bertsimas. Tractable stochastic analysis in high dimensions via robust optimization.Mathematical programming, 134(1):23–70, 2012

    2012

  16. [24]

    The dispersion bias.SIAM Journal on Financial Mathematics, 13(2):521–550, 2022

    Lisa R Goldberg, Alex Papanicolaou, and Alex Shkolnik. The dispersion bias.SIAM Journal on Financial Mathematics, 13(2):521–550, 2022

    2022

  17. [25]

    The quadratic optimization bias of large covariance matrices

    Hubeyb Gurdogan and Alex Shkolnik. The quadratic optimization bias of large covariance matrices. arXiv preprint arXiv:2410.03053, 2024

    2024 arXiv

  18. [26]

    Portfolio selection revisited.Annals of Operations Research, 346(1):137–155, 2025

    Alex Shkolnik, Alec Kercheval, Hubeyb Gurdogan, Lisa R Goldberg, and Haim Bar. Portfolio selection revisited.Annals of Operations Research, 346(1):137–155, 2025. 56

    2025

  19. [27]

    A tool for filtering information in complex systems.Proceedings of the National Academy of Sciences, 102(30):10421–10426, 2005

    Michele Tumminello, Tomaso Aste, Tiziana Di Matteo, and Rosario N Mantegna. A tool for filtering information in complex systems.Proceedings of the National Academy of Sciences, 102(30):10421–10426, 2005

    2005

  20. [28]

    Community characterization of heterogeneous complex systems.Journal of Statistical Mechanics: Theory and Experiment, 2011(01):P01019, 2011

    Michele Tumminello, Salvatore Micciche, Fabrizio Lillo, Jan Varho, Jyrki Piilo, and Rosario N Man- tegna. Community characterization of heterogeneous complex systems.Journal of Statistical Mechanics: Theory and Experiment, 2011(01):P01019, 2011

    2011

  21. [29]

    Relation between financial market structure and the real economy: comparison between clustering methods.PloS one, 10(3):e0116201, 2015

    Nicolo Musmeci, Tomaso Aste, and Tiziana Di Matteo. Relation between financial market structure and the real economy: comparison between clustering methods.PloS one, 10(3):e0116201, 2015

    2015

  22. [30]

    Robust optimization–methodology and applications.Mathe- matical Programming, 92(3):453–480, 2002

    Aharon Ben-Tal and Arkadi Nemirovski. Robust optimization–methodology and applications.Mathe- matical Programming, 92(3):453–480, 2002

    2002

  23. [31]

    Theory and applications of robust optimization.SIAM review, 53(3):464–501, 2011

    Dimitris Bertsimas, David B Brown, and Constantine Caramanis. Theory and applications of robust optimization.SIAM review, 53(3):464–501, 2011

    2011

  24. [32]

    Robust mean-covariance solutions for stochastic optimization.Operations Research, 55 (1):98–112, 2007

    Ioana Popescu. Robust mean-covariance solutions for stochastic optimization.Operations Research, 55 (1):98–112, 2007

    2007

  25. [33]

    Robust portfolio selection problems.Mathematics of operations research, 28(1):1–38, 2003

    Donald Goldfarb and Garud Iyengar. Robust portfolio selection problems.Mathematics of operations research, 28(1):1–38, 2003

    2003

  26. [34]

    Robust portfolio control with stochastic factor dynamics.Operations Research, 61(4):874–893, 2013

    Paul Glasserman and Xingbo Xu. Robust portfolio control with stochastic factor dynamics.Operations Research, 61(4):874–893, 2013

    2013

  27. [35]

    Tütüncü and Mark Koenig

    Reha H. Tütüncü and Mark Koenig. Robust asset allocation.Annals of Operations Research, 132(1): 157–187, 2004

    2004

  28. [36]

    Worst-case value-at-risk and robust portfolio optimization: A conic programming approach.Operations research, 51(4):543–556, 2003

    Laurent El Ghaoui, Maksim Oks, and Francois Oustry. Worst-case value-at-risk and robust portfolio optimization: A conic programming approach.Operations research, 51(4):543–556, 2003

    2003

  29. [37]

    Markowitz portfolio construction at seventy.arXiv preprint arXiv:2401.05080, 2024

    Stephen Boyd, Kasper Johansson, Ronald Kahn, Philipp Schiele, and Thomas Schmelzer. Markowitz portfolio construction at seventy.arXiv preprint arXiv:2401.05080, 2024

    2024 arXiv

  30. [38]

    Addressing estimation errors on expected asset returns through robust portfolio optimization.Quantitative Finance, pages 1–14, 2026

    Gérard Cornuéjols, Özgün Elçİ, and Vrishabh Patil. Addressing estimation errors on expected asset returns through robust portfolio optimization.Quantitative Finance, pages 1–14, 2026

    2026

  31. [39]

    Fabozzi, Dashan Huang, and Guofu Zhou

    Frank J. Fabozzi, Dashan Huang, and Guofu Zhou. Robust portfolios: Contributions from operations research and finance.Annals of Operations Research, 176(1):191–220, 2010. 57

    2010

  32. [40]

    Data-driven robust optimization.Mathematical Programming, 167(2):235–292, 2018

    Dimitris Bertsimas, Vishal Gupta, and Nathan Kallus. Data-driven robust optimization.Mathematical Programming, 167(2):235–292, 2018

    2018

  33. [41]

    Computing return estimation error matrices for robust optimization.Axioma Research Papers, 1:1–9, 2005

    RA Stubbs and P Vance. Computing return estimation error matrices for robust optimization.Axioma Research Papers, 1:1–9, 2005

    2005

  34. [42]

    Distributionally robust optimization under moment uncertainty with application to data-driven problems.Operations research, 58(3):595–612, 2010

    Erick Delage and Yinyu Ye. Distributionally robust optimization under moment uncertainty with application to data-driven problems.Operations research, 58(3):595–612, 2010

    2010

  35. [43]

    Incorporating asymmetric distributional information in robust value-at-risk optimization.Management Science, 54(3):573–585, 2008

    Karthik Natarajan, Dessislava Pachamanova, and Melvyn Sim. Incorporating asymmetric distributional information in robust value-at-risk optimization.Management Science, 54(3):573–585, 2008

    2008

  36. [44]

    Quantifying distributional model risk via optimal transport

    Jose Blanchet and Karthyek Murthy. Quantifying distributional model risk via optimal transport. Mathematics of Operations Research, 44(2):565–600, 2019

    2019

  37. [45]

    Distributionally robust stochastic optimization with wasserstein distance

    Rui Gao and Anton Kleywegt. Distributionally robust stochastic optimization with wasserstein distance. Mathematics of Operations Research, 48(2):603–655, 2023

    2023

  38. [46]

    Honey, I Shrunk the Sample Covariance Matrix.The Journal of Portfolio Management, 30(4):110–119, 2004

    Olivier Ledoit and Michael Wolf. Honey, I Shrunk the Sample Covariance Matrix.The Journal of Portfolio Management, 30(4):110–119, 2004

    2004

  39. [47]

    A test for the number of factors in an approximate factor model.the Journal of Finance, 48(4):1263–1291, 1993

    Gregory Connor and Robert A Korajczyk. A test for the number of factors in an approximate factor model.the Journal of Finance, 48(4):1263–1291, 1993

    1993

  40. [48]

    Estimation with quadratic loss

    William James, Charles Stein, et al. Estimation with quadratic loss. InProceedings of the fourth Berkeley symposium on mathematical statistics and probability, volume 1, pages 361–379. University of California Press, 1961

    1961

  41. [49]

    Awell-conditionedestimatorforlarge-dimensionalcovariancematrices

    OlivierLedoitandMichaelWolf. Awell-conditionedestimatorforlarge-dimensionalcovariancematrices. Journal of multivariate analysis, 88(2):365–411, 2004

    2004

  42. [50]

    Shrinkage algorithms for mmse covariance estimation.IEEE transactions on signal processing, 58(10):5016–5029, 2010

    Yilun Chen, Ami Wiesel, Yonina C Eldar, and Alfred O Hero. Shrinkage algorithms for mmse covariance estimation.IEEE transactions on signal processing, 58(10):5016–5029, 2010

    2010

  43. [51]

    Riskmetricstm—technical document.Morgan Guaranty Trust Company of New York: New York, 51:54, 1996

    Jacques Longerstaey and Martin Spencer. Riskmetricstm—technical document.Morgan Guaranty Trust Company of New York: New York, 51:54, 1996

    1996

  44. [52]

    Nonlinear shrinkage estimation of large-dimensional covariance ma- trices.The Annals of Statistics, pages 1024–1060, 2012

    Olivier Ledoit and Michael Wolf. Nonlinear shrinkage estimation of large-dimensional covariance ma- trices.The Annals of Statistics, pages 1024–1060, 2012. 58

    2012

  45. [53]

    Nonlinear shrinkage of the covariance matrix for portfolio selection: Markowitz meets goldilocks.The Review of Financial Studies, 30(12):4349–4388, 2017

    Olivier Ledoit and Michael Wolf. Nonlinear shrinkage of the covariance matrix for portfolio selection: Markowitz meets goldilocks.The Review of Financial Studies, 30(12):4349–4388, 2017

    2017

  46. [54]

    Analytical nonlinear shrinkage of large-dimensional covariance matri- ces.The Annals of Statistics, 48(5):3043–3065, 2020

    Olivier Ledoit and Michael Wolf. Analytical nonlinear shrinkage of large-dimensional covariance matri- ces.The Annals of Statistics, 48(5):3043–3065, 2020

    2020

  47. [55]

    A generalized approach to portfolio optimization: Improving performance by constraining portfolio norms.Management science, 55(5):798–812, 2009

    Victor DeMiguel, Lorenzo Garlappi, Francisco J Nogales, and Raman Uppal. A generalized approach to portfolio optimization: Improving performance by constraining portfolio norms.Management science, 55(5):798–812, 2009

    2009

  48. [56]

    Machine learning and portfolio optimization

    Gah-Yi Ban, Noureddine El Karoui, and Andrew EB Lim. Machine learning and portfolio optimization. Management Science, 64(3):1136–1154, 2018

    2018

  49. [57]

    A scalable algorithm for sparse portfolio selection.IN- FORMS Journal on Computing, 34(3):1489–1511, 2022

    Dimitris Bertsimas and Ryan Cory-Wright. A scalable algorithm for sparse portfolio selection.IN- FORMS Journal on Computing, 34(3):1489–1511, 2022

    2022

  50. [58]

    Inverse optimization: A new perspective on the black-litterman model.Operations research, 60(6):1389–1403, 2012

    Dimitris Bertsimas, Vishal Gupta, and Ioannis Ch Paschalidis. Inverse optimization: A new perspective on the black-litterman model.Operations research, 60(6):1389–1403, 2012

    2012

  51. [59]

    A review of two decades of correlations, hierarchies, networks and clustering in financial markets.Progress in information geometry: Theory and applications, pages 245–274, 2021

    Gautier Marti, Frank Nielsen, Mikołaj Bińkowski, and Philippe Donnat. A review of two decades of correlations, hierarchies, networks and clustering in financial markets.Progress in information geometry: Theory and applications, pages 245–274, 2021

    2021

  52. [60]

    Can machine learning-based portfolios outperform traditional risk-based portfolios? the need to account for covariance misspecification.Risks, 7(3):74, 2019

    Prayut Jain and Shashi Jain. Can machine learning-based portfolios outperform traditional risk-based portfolios? the need to account for covariance misspecification.Risks, 7(3):74, 2019

    2019

  53. [61]

    Cross asset portfolios of tradable risk premia indices.Global Quantitative and Derivatives Strategy, JP Morgan, 2017

    A Lau, M Kolanovic, T Lee, and R Krishnamachari. Cross asset portfolios of tradable risk premia indices.Global Quantitative and Derivatives Strategy, JP Morgan, 2017

    2017

  54. [62]

    The hierarchical equal risk contribution portfolio.Available at SSRN 3237540, 2018

    Thomas Raffinot. The hierarchical equal risk contribution portfolio.Available at SSRN 3237540, 2018

    2018

  55. [63]

    Hierarchical clustering-based asset allocation.Journal of portfolio management, 44 (2):89–99, 2018

    Thomas Raffinot. Hierarchical clustering-based asset allocation.Journal of portfolio management, 44 (2):89–99, 2018

    2018

  56. [64]

    Hierarchical structure in financial markets.The European Physical Journal B- Condensed Matter and Complex Systems, 11(1):193–197, 1999

    Rosario N Mantegna. Hierarchical structure in financial markets.The European Physical Journal B- Condensed Matter and Complex Systems, 11(1):193–197, 1999

    1999

  57. [65]

    Common risk factors in the returns on stocks and bonds

    Eugene F Fama and Kenneth R French. Common risk factors in the returns on stocks and bonds. Journal of financial economics, 33(1):3–56, 1993. 59

    1993

  58. [66]

    Returns to buying winners and selling losers: Implications for stock market efficiency.The Journal of finance, 48(1):65–91, 1993

    Narasimhan Jegadeesh and Sheridan Titman. Returns to buying winners and selling losers: Implications for stock market efficiency.The Journal of finance, 48(1):65–91, 1993

    1993

  59. [67]

    Value and momentum everywhere

    Clifford S Asness, Tobias J Moskowitz, and Lasse Heje Pedersen. Value and momentum everywhere. The journal of finance, 68(3):929–985, 2013

    2013

  60. [68]

    Time series momentum.Journal of financial economics, 104(2):228–250, 2012

    Tobias J Moskowitz, Yao Hua Ooi, and Lasse Heje Pedersen. Time series momentum.Journal of financial economics, 104(2):228–250, 2012

    2012

  61. [69]

    Springer, 2021

    Michael L Bynum, Gabriel A Hackebeil, William E Hart, Carl D Laird, Bethany L Nicholson, John D Siirola, Jean-Paul Watson, David L Woodruff, et al.Pyomo-optimization modeling in python, volume 67. Springer, 2021

    2021

  62. [70]

    On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming.Mathematical programming, 106(1):25–57, 2006

    Andreas Wächter and Lorenz T Biegler. On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming.Mathematical programming, 106(1):25–57, 2006. 60 Supplementary Material S1 Extended Results S1.1 Rank Heatmaps: Calibrated Sweep ...

    2006

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.