Pith. sign in

REVIEW 2 major objections 5 minor 1 cited by

Herding and Liquidity in Order-Book Markets. I. A Robust Liquidity-Stress Crossover and its Reflexive Mechanism

T0 review · 2 major / 5 minor · reviewed 2026-07-13 · grok-4.5

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

Pith's one-line read Herding produces a smooth, null-verified liquidity dry-up in an order book, driven by a reflexive price-momentum loop, not a discontinuous crash.

desk verdict Solid, disciplined phase-diagram work on order-book herding: a null-verified liquidity dry-up crossover with a signed reflexive mechanism under momentum, not a Dark Corner. read the letter →

arxiv 2607.08907 v1 pith:U4HGVF5N submitted 2026-07-09 q-fin.TR q-fin.CP

Jan Novotny This is my paper · ORCID
classification q-fin.TRq-fin.CP
keywords agent-basedmodellingmarketmicrostructureherdingliquiditycrisesreflexivityorderbookphasediagram
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

Agent-based market models often throw up apparent instabilities that may be nothing more than parameter quirks. This paper maps the full two-dimensional phase diagram of a continuous double-auction order book whose agents mix fundamental-anchored zero-intelligence liquidity with mid-anchored chartist herders, controlled by herd fraction and herd strength. It finds a genuine liquidity-stress crossover: when both controls are high the book is one-sided roughly a third of the time. The dry-up vanishes completely under a scrambled-sign null that keeps the same order mix but destroys directional correlation, so the effect is imitation-specific rather than a mere thinning of liquidity. The same corner stress reappears under an order-flow-imbalance herding rule, survives a sixteen-fold change of momentum horizon, and has a clean monotone onset boundary that falls as herding strength rises. Mechanism decomposition at matched directional-bias amplitude shows that price-momentum herding carries a large, comparator-robust reflexive component—buying begets buying—while the order-flow rule does not. The result is a characterised liquidity phenomenon, not a price bubble or a sharp tipping line, and a companion two-market test finds no directional contagion across a pure signal link.

What carries the argument

The scrambled-sign null together with amplitude-matched open-loop shadow/replay decomposition: the null randomises only herd buy/sell signs while preserving order-type mix and zero-intelligence dilution, isolating directional imitation; the matched-amplitude comparison then measures how much extra dry-up the closed loop produces over equally forceful open-loop drives, revealing a robust self-reinforcing component under price momentum.

What would settle it

Re-run the high-herding corner with a control that preserves higher-moment burst structure of the herd signal while still destroying directional correlation; if the one-sided-book fraction remains high under that control, the imitation-specific claim fails.

Watch

Extended reading notes

Core claim

A null-verified, imitation-specific liquidity-stress crossover sits in the high-herding corner of the herd-fraction × herd-strength plane. The fraction of one-sided-book events rises to about 0.34 at the corner, is identically zero across the entire scrambled-sign null plane, is rule- and horizon-robust, forms a smooth crossover rather than a discontinuous Dark Corner, and under price-momentum herding is amplified by a large comparator-robust reflexive component of roughly +0.29 at matched directional-bias amplitude.

Load-bearing premise

The scrambled-sign null is assumed to isolate directional imitation as the sole cause of the dry-up, so that any real-minus-null difference is truly due to correlated herding rather than residual timing or burst properties of the closed-loop signal.

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

2 major / 5 minor

Summary. The paper applies Bouchaud's phase-diagram discipline to a continuous double-auction order-book ABM with fundamental-anchored zero-intelligence liquidity and a mid-anchored chartist herding layer controlled by herd fraction φ and strength κ. A 7×6 grid (336 runs with scrambled-sign nulls) locates an emergent liquidity-stress crossover: the one-sided-book fraction ϕ∅ rises to ≈0.34 at (φ,κ)=(0.9,1.0), is identically zero across all 42 scrambled cells, and forms a smooth ramp rather than a discontinuous Dark Corner. The dry-up is rule-robust under an order-flow-imbalance signal, horizon-robust across a 16× momentum-window range, and has a monotone onset boundary φ*(κ)={0.55,0.45,0.36}. At matched directional-bias amplitude (mean |p_buy−0.5|≈0.269), price-momentum herding carries a large comparator-robust reflexive component (+0.29), while the OFI component is ≈0 and comparator-dependent. RMS-mispricing growth is identified as a placement artefact (largest at κ=0), and a companion two-market analysis finds no directional contagion across a signal-only herding link.

Significance. If the result holds, this is a substantial methodological and substantive contribution to market-microstructure ABMs. It is, to the authors' knowledge and mine, the first full application of Bouchaud's phase-diagram / artefact-ruling-out programme to an order-book vehicle with herding and liquidity as control axes. The scrambled-sign null, band and numerical-stability controls, rule- and horizon-robustness battery, fine onset scans, and amplitude-matched multi-comparator open-loop decomposition (shadow, replay, synthetic telegraph) are unusually thorough and reproducible; code and data are public. The signed, rule-dependent reflexive mechanism and the no-contagion corollary are falsifiable predictions that distinguish the work from exogenous-shock flash-crash calibrations and from sharp Dark-Corner claims in the macro lineage. The careful refusal to over-claim a tipping line is itself a strength.

major comments (2)
  1. §5.6 / Table 7: The OFI reflexive component is reported as ≈0 and comparator-dependent, yet the headline contrast with momentum's +0.29 is load-bearing for the 'signed, rule-dependent' mechanism claim. The baseline-flow shadow (0.418) is a ~2–3× outlier among matched-amplitude drives and rests on only 3 independent driving series; the paper itself notes that seed-level SEM understates uncertainty and that clustering widens the interval. A higher-moment-matched open-loop comparator (or a larger driving-series ensemble with clustered errors) is needed before the asymmetry can be stated as firmly as in the abstract and §6.2. Without it the OFI side remains under-resolved rather than cleanly unsigned.
  2. §3.4 and §5.1: The scrambled-sign null is the central causal instrument for the 'imitation-specific' claim. It randomises only buy/sell signs while preserving order-type mix and φ-driven ZI dilution. Residual non-sign structure of real herding (timing, burstiness, mid-anchoring interaction with the closed-loop signal) is partially isolated by the later shadow/replay/telegraph legs, but those legs are run only at the corner and only after amplitude matching. A brief explicit check that null and real share comparable higher-moment order-arrival statistics (or a null that also scrambles arrival timing) would close the residual channel the authors themselves flag as the weakest premise.
minor comments (5)
  1. §3.5: The continuous-reference convention (carry forward last traded price when mid is undefined) is essential for fair real-vs-null comparison; it should be stated once in the main text with a short sensitivity note, not only in the methods paragraph.
  2. Figure 1 / Table 1: The tabulated zeros for low-φ cells are roundings; the fine-scan nonzero tail (ϕ∅=0.002 at (0.30,1.0)) is mentioned in text but not shown. A footnote or inset would prevent a reader from misreading a hard threshold.
  3. §5.7: The two-market no-contagion result is a clean corollary; the coupling form and the intensity-modulation variant are described briefly. A one-sentence statement of the exact signal-injection equation would aid replication.
  4. Notation: the order parameter is written both ϕ∅ and ϕ∅ (and occasionally ϕ∅) inconsistently across abstract, tables and body; standardise.
  5. Limitations §6.4 correctly note that ϕ∅ is still rising at the corner and that φ o1 is deliberately avoided; a single sentence in the conclusion reminding the reader that the map does not reach a plateau would keep the scope clear.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: order parameter, null contrast, and reflexive component are operational simulation measurements against independent open-loop and scrambled controls, not tautologies forced by definition or self-citation.

full rationale

The paper maps a two-dimensional phase diagram of a continuous-double-auction ABM by direct Monte-Carlo sweeps (7 imes6 grid plus robustness batteries) and reports measured values of the order parameter ϕ∅ (fraction of one-sided-book events). ϕ∅ is defined as an observable computed from the book state, not in terms of the claimed crossover; the scrambled-sign null (identical order-type mix and φ-driven ZI dilution, randomized signs only) is an independent control condition that yields identically zero ϕ∅ across all cells, licensing the imitation-specific reading without circular reduction. The reflexive component is likewise an operational difference (closed-loop real minus amplitude-matched open-loop shadow/replay/telegraph drives at mean |p_buy−0.5|≈0.269), obtained by separate simulation legs rather than by algebraic identity or fitted-parameter renaming. Amplitude matching is a design control that equalizes first-moment force before comparison; it does not force the subsequent excess (+0.29 under momentum). Onset boundary φ*(κ) is extracted by linear interpolation of measured ϕ∅ curves, not presupposed. Citations (Bouchaud manifesto, Lux–Marchesi, Cont–Bouchaud, Fosset et al.) supply methodological precedent and classical herding ingredients; none is a self-citation that supplies a uniqueness theorem or ansatz whose content is the present result. No equation equates a claimed prediction to its own input by construction, and the findings remain falsifiable inside the model class by the reported nulls and comparators. The derivation chain is therefore self-contained computational measurement, not circular.

Assumptions & free parameters 7 free parameters · 5 assumptions · 2 invented entities

The central claim is a simulation finding inside a constructed market vehicle. It rests on standard continuous-double-auction and zero-intelligence modeling assumptions, a specific mid-anchored herding response, hand-chosen microstructure parameters, and operational definitions of the order parameter and reflexive component. No new physical entity is postulated; free parameters are experimental controls and baseline ZI settings rather than fits to market data. The ledger is therefore mostly domain modeling choices plus a few ad-hoc operational conventions (onset threshold, matched amplitude target).

free parameters (7)
  • ZI market-order probability p_market
    Baseline microstructure parameter fixed at 0.15; shapes book dynamics and how fast sides can be consumed.
  • ZI cancellation probability p_cancel
    Fixed at 0.05; controls resting liquidity lifetime and replenishment.
  • limit-order placement bandwidth b
    Default b=4 (swept in {2,4,8} for robustness); sets scatter of resting liquidity and mildly shifts onset.
  • fundamental volatility σ_f
    Fixed at 0.05; keeps price near fundamental while stress is measured in sidedness.
  • momentum window w and tanh response scale
    Defaults w=200, scale=2.0 (both swept); define herd signal strength and horizon; OFI scale bisected to 68.8 for amplitude match.
  • matched directional-bias amplitude mean |p_buy−0.5|≈0.269
    Chosen as momentum’s default-scale operating point and used as the fair-comparison target for open-loop decomposition; not a market-data fit but a design choice that conditions the reflexive-component numbers.
  • onset threshold ϕ∅=0.02 for φ*(κ)
    Convention for extracting the onset boundary; authors show monotone shape across 0.005–0.05 but absolute φ* shifts.
assumptions (5)
  • domain assumption Continuous double-auction limit-order matching with fundamental-anchored zero-intelligence liquidity is a valid baseline market vehicle.
    §3.1–3.6: foundation validation at (φ,κ)=(0,0) and claim that both stable and stressed regimes are reachable.
  • ad hoc to paper Herders place mid-anchored orders whose side follows p_buy=0.5+0.5κ tanh(signal/scale) for momentum or OFI signals.
    §3.2–3.3 and §5.6: defines the herding layer and the nonlinearity that forces amplitude matching.
  • ad hoc to paper Scrambling only herd order signs while preserving order-type mix and φ-driven ZI dilution isolates directional imitation.
    §3.4: central causal identification strategy for calling the dry-up imitation-specific.
  • ad hoc to paper When the mid is undefined, carrying forward the last traded price yields a consistent order-parameter definition across real and null conditions.
    §3.5: avoids selection bias that would drop stressed events only in the real condition.
  • domain assumption Bouchaud phase-diagram discipline (map controls, null, robustness, artefact checks) is the right standard for calling a feature emergent rather than artefactual.
    §1–2: methodological frame imported from Mark-0 / Dark Corners literature.
invented entities (2)
  • Order parameter ϕ∅ (fraction of events with a one-sided book)
    purpose: Separates stressed from quiescent regimes and serves as the primary observable of the liquidity-stress crossover.
    Defined operationally in §3.5; standard as an order parameter concept but specific to this paper’s sidedness definition and continuous-reference fallback.
  • Reflexive component (closed-loop real minus amplitude-matched open-loop drive)
    purpose: Quantifies extra dry-up manufactured by the herding loop beyond equally forceful exogenous drive.
    §4.2 and §5.6: constructed measurement, not an external market observable with independent empirical handle in this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Herding and Liquidity in Order-Book Markets. I. A Robust Liquidity-Stress Crossover and its Reflexive Mechanism." pith.science (2026). https://pith.science/paper/U4HGVF5N

@misc{pith2026260708907,
  author       = {Pith},
  title        = {Pith review of: Herding and Liquidity in Order-Book Markets. I. A Robust Liquidity-Stress Crossover and its Reflexive Mechanism},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/U4HGVF5N}},
  note         = {Machine review of arXiv:2607.08907}
}
abstract

Agent-based models of markets readily produce emergent instabilities, but telling a genuine collective effect apart from a parameter artefact takes discipline. We apply Bouchaud's phase-diagram method to a continuous-double-auction order-book model. The method is to map the full phase diagram, test its robustness to rule changes, and rule out degenerate and numerical origins before we call any feature a tipping point. The model has fundamental-anchored zero-intelligence liquidity and a mid-anchored chartist herding layer, controlled by the fraction $\varphi$ and the strength $\kappa$ of herders. A 7x6 grid (336 runs, each with a scrambled-sign null) locates an emergent liquidity-stress crossover. The order parameter, the fraction of events with a one-sided book, rises to about 0.34 at $(\varphi,\kappa)=(0.9,1.0)$, is zero across all 42 scrambled cells, and forms a smooth crossover rather than a discontinuous Dark Corner. The dry-up is rule-robust (it recurs under an order-flow-imbalance rule), horizon-robust (about 0.32-0.35 across a 16x range of momentum window), and has a monotone onset boundary $\varphi^*(\kappa) = \{0.55, 0.45, 0.36\}$. We then decompose the mechanism at a matched directional-bias amplitude (mean |p_buy - 0.5| about 0.269). Price-momentum herding carries a large, comparator-robust reflexive component (+0.29; buying begets buying), whereas the order-flow rule's component is about 0 and comparator-dependent. The RMS-mispricing gradient is a placement artefact, largest at $\kappa=0$. A companion two-market analysis finds no directional cross-market contagion across a signal-only herding link.

Figures

Figures reproduced from arXiv: 2607.08907 by the authors.

Figure 1
Figure 1. Order parameters, REAL herding (top) vs. SCRAMBLED-sign null (bottom), shared [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Cut across the high-herding corner at κ = 1.0, real vs. scrambled null (mean±SEM). Right: ϕ∅ rises smoothly and monotonically for the real condition (blue) from φ ≈ 0.3 onward, saturating at ≈ 0.34; the null (orange) is flat at zero–a smooth, imitation-specific crossover. Left: RMS mispricing–the null (orange) rises to ≈ 6.7 at φ = 0.9 while the real curve (blue) stays flat at ≈ 2.5, showing that directional imitati… view at source ↗
Figure 3
Figure 3. Rule-robustness under the OFI herding signal. [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Horizon-robustness (price-momentum signal), corner (0 [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Onset boundary of the liquidity-stress region (scrambled null = 0 at all cells). [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Momentum has a robust self-reinforcing reflexive component; OFI’s is comparator [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: No cross-market contagion (decisive fair test). With B near its own onset and A’s push [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: Amplitude × persistence continuum of the dry-up under a synthetic telegraph drive at the corner. ϕ∅ increases smoothly and monotonically along both axes, with no threshold. Momentum’s and matched-OFI’s real operating points are overlaid; the OFI operating point’s actua…
Figure 9
Figure 9. Figure 9: Channel decomposition, REAL − SCRAMBLED for RMS mispricing. The imitation channel does not add mispricing over the placement/dilution baseline captured by the null; consis￾tent with the reading that the RMS gradient is a placement artefact. 5.10 Other order parameters …
Figure 10
Figure 10. Figure 10: All six order parameters across the φ×κ plane (REAL condition). Only ϕ∅ (the liquidity dry-up) shows a clean, localised region; the mispricing, excursion, drawdown and volatility panels show no tipping structure. orders become correlated. They then repeatedly consume …

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

  1. Herding and Liquidity in Order-Book Markets. II. Fundamental Anchoring and the Resilience of Liquidity

    q-fin.TR 2026-07 conditional novelty 6.0 of 10

    In a two-market order-book agent-based model, fundamental anchoring is the stabilizer whose removal lets a leverage spiral self-sustain, while none of six coupling channels transmits liquidity stress between markets.

Reference graph

Works this paper leans on

18 extracted references · 5 canonical work pages · cited by 1 Pith paper

  1. [1]

    Navigating through economic complexity: Phase diagrams & pa- rameter sloppiness

    Jean-Philippe Bouchaud. Navigating through economic complexity: Phase diagrams & pa- rameter sloppiness. arXiv:2412.11259, 2024

    arXiv 2024

  2. [2]

    Monetary policy and dark corners in a stylized agent-based model

    Stanislao Gualdi, Marco Tarzia, Francesco Zamponi, and Jean-Philippe Bouchaud. Monetary policy and dark corners in a stylized agent-based model. arXiv:1501.00434, 2015

    arXiv 2015

  3. [3]

    Tipping points in macroeconomic agent-based models.Journal of Economic Dynamics & Control, 50: 29–61, 2015

    Stanislao Gualdi, Marco Tarzia, Francesco Zamponi, and Jean-Philippe Bouchaud. Tipping points in macroeconomic agent-based models.Journal of Economic Dynamics & Control, 50: 29–61, 2015. doi: 10.1016/j.jedc.2014.08.003

    work page doi:10.1016/j.jedc.2014.08.003 2015

  4. [4]

    Emergent inequalities in a primitive agent-based good-exchange model

    Nirbhay Patil and Jean-Philippe Bouchaud. Emergent inequalities in a primitive agent-based good-exchange model. arXiv:2405.18116, 2024

    arXiv 2024

  5. [5]

    Understanding agent-based models of financial markets: A bottom-up approach based on order parameters and phase diagrams

    Ribin Lye, James Peng Lung Tan, and Siew Ann Cheong. Understanding agent-based models of financial markets: A bottom-up approach based on order parameters and phase diagrams. Physica A: Statistical Mechanics and its Applications, 391(21):5521–5531, 2012

    2012

  6. [6]

    Endogenous liquidity crises

    Amaury Fosset, Jean-Philippe Bouchaud, and Michael Benzaquen. Endogenous liquidity crises. Journal of Statistical Mechanics: Theory and Experiment, 2020(6):063401, 2020. doi: 10.1088/ 1742-5468/ab7c64

    2020

  7. [7]

    Kang Gao, Perukrishnen Vytelingum, Stephen Weston, Wayne Luk, and Ce Guo. High- frequency financial market simulation and flash crash scenarios analysis: An agent-based modelling approach.Journal of Artificial Societies and Social Simulation, 27(2):8, 2024. doi: 10.18564/jasss.5403

    work page doi:10.18564/jasss.5403 2024

  8. [8]

    Scaling and criticality in a stochastic multi-agent model of a financial market.Nature, 397:498–500, 1999

    Thomas Lux and Michele Marchesi. Scaling and criticality in a stochastic multi-agent model of a financial market.Nature, 397:498–500, 1999. doi: 10.1038/17290

    doi:10.1038/17290 1999

Show all 18 references
  1. [9]

    Herd behavior and aggregate fluctuations in financial markets.Macroeconomic Dynamics, 4(2):170–196, 2000

    Rama Cont and Jean-Philippe Bouchaud. Herd behavior and aggregate fluctuations in financial markets.Macroeconomic Dynamics, 4(2):170–196, 2000. doi: 10.1017/S1365100500015029. 25

    doi

  2. [10]

    Continuum time limit and stationary states of the mi- nority game.Physical Review E, 64:056138, 2001

    Matteo Marsili and Damien Challet. Continuum time limit and stationary states of the mi- nority game.Physical Review E, 64:056138, 2001. doi: 10.1103/PhysRevE.64.056138

    2001 doi

  3. [11]

    Quantifying reflexivity in financial markets: Towards a prediction of flash crashes.Physical Review E, 85:056108, 2012

    Vladimir Filimonov and Didier Sornette. Quantifying reflexivity in financial markets: Towards a prediction of flash crashes.Physical Review E, 85:056108, 2012. doi: 10.1103/PhysRevE.85. 056108

    2012 doi

  4. [12]

    The deeds of speed: An agent-based model of market liquidity and flash episodes

    Geir-Are K˚ arvik, Joseph Noss, Jack Worlidge, and Daniel Beale. The deeds of speed: An agent-based model of market liquidity and flash episodes. Staff Working Paper 743, Bank of England, 2018. URLhttps://www.bankofengland.co.uk/working-paper/2018/ the-deeds-of-speed-an-agent-...

    2018

  5. [13]

    Waterfall, Fergal P

    Joshua J. Waterfall, Fergal P. Casey, Ryan N. Gutenkunst, Kevin S. Brown, Christopher R. Myers, Piet W. Brouwer, Veit Elser, and James P. Sethna. The sloppy model universality class and the vandermonde matrix.Physical Review Letters, 97:150601, 2006. doi: 10.1103/ PhysRevLett....

    2006

  6. [14]

    Understanding flash crash con- tagion and systemic risk: A micro-macro agent-based approach

    James Paulin, Anisoara Calinescu, and Michael Wooldridge. Understanding flash crash con- tagion and systemic risk: A micro-macro agent-based approach. arXiv:1805.08454, 2018

    2018 arXiv

  7. [15]

    The fused asset flow model: Stability, bifurcation, and contagion in multi-asset markets with heterogeneous investors

    Mario Cavani. The fused asset flow model: Stability, bifurcation, and contagion in multi-asset markets with heterogeneous investors. arXiv:2605.28417, 2026

    2026 arXiv

  8. [16]

    Illiquidity contagion and liquidity crashes.Review of Financial Studies, 27(6):1615–1660, 2014

    Giovanni Cespa and Thierry Foucault. Illiquidity contagion and liquidity crashes.Review of Financial Studies, 27(6):1615–1660, 2014. doi: 10.1093/rfs/hhu016

    2014 doi

  9. [17]

    Running for the exit: Distressed selling and endogenous correlation in financial markets.Mathematical Finance, 23(4):718–741, 2013

    Rama Cont and Lakshithe Wagalath. Running for the exit: Distressed selling and endogenous correlation in financial markets.Mathematical Finance, 23(4):718–741, 2013. doi: 10.1111/j. 1467-9965.2011.00510.x

    2013 doi

  10. [18]

    Dis- secting cross-impact on stock markets: An empirical analysis.Journal of Statistical Mechanics: Theory and Experiment, 2017(2):023406, 2017

    Michael Benzaquen, Iacopo Mastromatteo, Zoltan Eisler, and Jean-Philippe Bouchaud. Dis- secting cross-impact on stock markets: An empirical analysis.Journal of Statistical Mechanics: Theory and Experiment, 2017(2):023406, 2017. doi: 10.1088/1742-5468/aa53f7. 26

    2017 doi

Pith tools

Reviewed July 13, 2026 · model on record in the stance chip above.