Optimal Signal Extraction from Order Flow: A Matched Filter Perspective on Normalization and Market Microstructure

Battle Count: 71

Highly relevant to quantitative trading. The paper provides a theoretically grounded and empirically validated framework for choosing the correct normalization when constructing order flow signals for alpha generation. Key practical implications: (1) Use SMC=D/M for domestic institutional flow signals, (2) Use STV=D/V for foreign/algorithmic flow signals, (3) The Participation Rate Fallacy warns against misinterpreting volume-normalized signals, (4) The Informed Executor Hypothesis suggests foreign volume-scaled flow carries strong predictive information (t=16.35) that is missed with wrong normalization, (5) Long-short portfolio results show SMC achieves Sharpe 5.48 vs STV 3.29 for domestic institutions, and STV_foreign achieves Sharpe 7.53. Directly applicable to factor construction, signal generation, and execution algorithm design.

Moderate complexity. The core methodology (normalizing order flow by market cap vs trading value, running Fama-MacBeth regressions, horse race specifications) is straightforward for practitioners familiar with cross-sectional return prediction. The theoretical matched filter derivation requires signal processing background. Key implementation challenges: (1) obtaining investor-type-disaggregated order flow data (available in Korea via KRX, harder in US), (2) proper standardization within each day, (3) interpreting horse race sign reversals correctly, (4) horizon analysis to separate information from price impact. The Monte Carlo simulation framework is well-specified and reproducible.

Theoretical framework is fully specified with mathematical derivations in Appendix A. Monte Carlo simulation parameters are detailed in Table 2. However, empirical data requires Korea Exchange (KRX) and CREON subscriptions. No code repository is provided. The Fama-MacBeth methodology and horse race regression specifications are clearly described, enabling replication given data access.