Model-Free Passive Execution via Order-Level Shadowing

Battle Count: 65

Directly addresses algorithmic execution, a core problem in quantitative trading. Proposes a model-free benchmark for passive order placement that execution desks can use to evaluate predictive placement models. The latency sensitivity analysis and adverse selection findings are practically relevant for HFT and execution algorithm design. The identifier-driven cancel mirroring mechanism offers a novel approach to order lifecycle management. However, it is a benchmark/reference method rather than a profit-generating strategy.

The core algorithm is specified in a single page (Algorithm 1) and requires only: (1) consuming order-book messages, (2) placing a limit order at the observed add's price/venue, (3) recording a single identifier association, (4) cancelling on delete/trade of the followed order. No order-book model, no fill prediction, no routing optimization, no state-dependent cancellation logic. The reference implementation adds safety guards (hysteresis bands, completion/over-hedge guards) but these are straightforward risk limits. The main complexity lies in the infrastructure (MBO book reconstruction, deterministic replay simulator) rather than the algorithm itself.

Fully open-source implementation (kaspar-hft in C++), deterministic replay simulator, specific parameters provided (placement rates, grace windows, band widths), full algorithm pseudocode in Algorithm 1, fixed session replays to identical fills. GitHub repository publicly available. Evaluation uses publicly available CME market-by-order data (channel 310).