Aliens on Earth: Early Detection of Latent Microstructure Regimes in Limit Order Books

Battle Count: 79

Highly relevant to quantitative trading. The paper directly addresses early detection of liquidity stress in limit order books, which is critical for market-making strategies, algorithmic execution, and risk management. The finding that depth erosion (not imbalance or volatility) is the leading indicator has direct practical implications for HFT firms. The theoretical guarantees provide principled bounds on achievable lead-time. The adaptive threshold mechanism is deployable in real-time streaming settings. However, the small real-data sample (5 events) and simulation-heavy evaluation temper immediate practical applicability.

Moderate complexity. The core algorithm (Algorithm 1) is straightforward: HMM forward filtering, four channel computations, MAX aggregation, rising-edge check, and adaptive threshold update—all O(1) per timestep after burn-in. However, proper implementation requires careful handling of: (1) causal z-scoring without look-ahead, (2) intraday seasonality removal, (3) rolling HMM re-fitting with multiple restarts, (4) correct CUSUM stopping time vs. partial sum distinction, and (5) adaptive threshold calibration. The theoretical proofs are mathematically sophisticated but not required for implementation.

Code and data are publicly available on GitHub and Zenodo (DOI: 10.5281/zenodo.19697687). The paper provides complete algorithmic pseudocode (Algorithm 1 and 2), a Python implementation sketch in Appendix F, full simulation parameters, and rigorous mathematical proofs. However, the real-data application uses only 5 events from Binance WebSocket feed, limiting full reproducibility of empirical claims. The simulation setup is fully specified with all DGP parameters.