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

Battle Count: 78

Highly relevant for understanding endogenous liquidity stress mechanisms in order-book markets. The finding that price-momentum herding creates a robust self-reinforcing reflexive loop (buying begets buying, ~7× more dry-up than exogenous drive) directly informs algorithmic trading strategy design, liquidity provision risk, and flash-crash early warning. The no-contagion result (stress is intrinsically local under signal-only herding) is relevant for multi-asset portfolio risk. However, the paper is primarily a methodological/foundational study rather than a direct trading strategy paper. The smooth crossover (not sharp tipping point) finding tempers urgency but the onset boundary φ*(κ) provides actionable thresholds for liquidity stress monitoring.

The agent-based model itself (CDA order book with ZI liquidity and herding layer) is moderately complex but well-specified. The full experimental battery (336+ base runs, ~320 robustness runs, amplitude-matched decomposition with multiple comparators, synthetic telegraph drives, two-market coupling) requires significant computational resources and careful experimental design. The phase-diagram methodology with scrambled-sign nulls, bandwidth robustness cuts, numerical-stability audits, and multi-comparator reflexive decomposition adds substantial methodological complexity. Reproducing the full paper requires implementing the CDA engine, two herding rules, the null protocol, and all robustness/mechanism experiments.

Code and data are publicly available at https://github.com/hanssmail/darkcorners-abm. The paper provides extremely detailed simulation parameters (grid dimensions, seed counts, event counts, burn-in periods, all parameter values). Unique never-reused seeds per condition, no common random numbers. All 336 base runs and ~320 robustness runs are fully specified. The methodology is transparent with explicit null controls and robustness batteries.