Realtime price impact detection

Battle Count: 62

Highly relevant to algorithmic execution and market microstructure. Addresses the central operational problem of detecting market impact in real-time during order execution. Directly applicable to execution algorithm design, venue selection, and dynamic trading adjustments. The causal framing (distinguishing self-caused impact from alpha/competition) is a key practical concern for algo traders. The method's ability to react within a few fills rather than hundreds makes it operationally viable for live execution.

The core algorithm is relatively straightforward: estimate local Poisson intensity from recent durations, compute exponential/Lomax p-values for fill-to-print durations, and combine via Fisher's method (closed-form chi-squared tail). The mathematical machinery is well-specified with closed-form solutions. However, practical implementation requires careful handling of: (1) adverse-event selection (filtering prints by direction), (2) window size selection for local rate estimation, (3) handling of non-Poisson clustering in real markets, (4) real-time event stream processing, and (5) integration with existing execution algorithms. The Hawkes extension would significantly increase complexity.

The paper provides complete mathematical derivations (Bayesian and frequentist predictive densities, Fisher's method combination, Poisson process formulation) and describes the Monte Carlo simulation methodology. However, no code repository is provided, and the simulation parameters (surprise levels, window sizes) are described only qualitatively via Figure 2. The economic validation requires proprietary real execution data not available to readers. The statistical machinery is fully specified and reproducible from the equations.